CCRVB-18: Felber, DNA/LION MSD and cell populations

### get tab names
xlsx_file_path <- file.path(input_dir,"raw_datasets.xlsx")
tab_names<-openxlsx::getSheetNames(xlsx_file_path)
### list with separate element for each tab
list_of_datasets <- lapply(tab_names,
                           openxlsx::read.xlsx,xlsxFile=xlsx_file_path)
names(list_of_datasets) <- tab_names

### open list for data plotted in each figure panel
list_of_diff_counts <- list()
list_of_stats <- list()

Figure 1

Fig 1B

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure1BDEF"]] %>%
  dplyr::filter(Fig %in% c("figure 1B upper panel","figure 1B bottom panel")) %>%
  ### add log10
  dplyr::mutate(log10_value = log10(value_raw + prior_to_use))

colors_df <- data.frame("panel"=c("figure 1B upper panel","figure 1B bottom panel"),
                        "color"=c("darkred","grey20"))


### To name output files
contrast_name <- "Fig1B"

num_columns_for_plot <- 1

list_of_plots_mean <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i) %>%
   ## make timepoint for plotting
   dplyr::mutate(TimePoint = factor(TimePoint,unique(dplyr::filter(data_to_use, Fig == i)$TimePoint))) 
                   
  curr_color <- colors_df$color[which(colors_df$panel == i)]


 
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = log10_value, group = MacaqueID)) + 
    geom_line(color = curr_color, linewidth = 0.3)+
    scale_x_discrete(labels = unique(phen_to_plot$TimePoint) %>%
                       str_replace("w",", w") %>%
                       str_remove_all("k"))+
    #scale_y_continuous(transform = "log10")+
    labs(title = i)+
    ylab("log10(value)")+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_blank(), 
            axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

   p_mean <- p +
    stat_summary(aes(group = 1, y = log10_value), fun=mean, geom="line", colour=curr_color, linewidth = 1)+
  
    stat_summary(aes(group = 1, y = log10_value), fun=mean, geom="point", colour=curr_color)

      # plot(p)

    list_of_plots_mean[[i]] <- p_mean   
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots_mean, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig 1D

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure1BDEF"]] %>%
  dplyr::filter(Fig == c("figure1D")) %>%
  ### add log10
  dplyr::mutate(log10_value = log10(value_raw + prior_to_use),
                TimePoint = as.numeric(TimePoint))

#### set y axis limits
y_axis_limits <- c(1,6)
shapes_to_use <- c(24,21,4,22,25)


### To name output files
contrast_name <- "Fig1D"

num_columns_for_plot <- 1

  p <- ggplot(data_to_use, aes(x = TimePoint, y = log10_value, group = MacaqueID)) + 
    geom_line(color = "darkblue", linewidth = 0.3)+
  geom_point(aes(shape = MacaqueID), color = "darkblue", fill="darkblue")+
    scale_x_continuous(limits = c(0,120), breaks = c(seq(0,28,4),seq(40,120,20)))+
    scale_y_continuous(limits = y_axis_limits)+
  scale_shape_manual(values = shapes_to_use)+
    labs(title = unique(data_to_use$Fig))+
  xlab("weeks of study")+
  ylab("log10(value)")+
  theme_bw()+
        theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"))
plot(p)

Fig 1E

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure1BDEF"]] %>%
  dplyr::filter(Fig == c("figure1E")) %>%
  ### add log10
  dplyr::mutate(log10_value = value_raw, ### already logged
                TimePoint = as.numeric(TimePoint))

#### set y axis limits
y_axis_limits <- c(0.5,5)
shapes_to_use <- c(24,21,22,25)

### To name output files
contrast_name <- "Fig1E"

num_columns_for_plot <- 1

  p <- ggplot(data_to_use, aes(x = TimePoint, y = log10_value, group = MacaqueID)) + 
    geom_line(color = "darkblue", linewidth = 0.3)+
  geom_point(aes(shape = MacaqueID), color = "darkblue", fill="darkblue")+
    scale_x_continuous(limits = c(0,120), breaks = c(seq(0,28,4),seq(40,120,20)))+
    scale_y_continuous(limits = y_axis_limits)+
  scale_shape_manual(values = shapes_to_use)+
    labs(title = unique(data_to_use$Fig))+
  xlab("weeks of study")+
  ylab("log10(value)")+
  theme_bw()+
        theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"))
plot(p)

Fig 1F

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure1BDEF"]] %>%
  dplyr::filter(Fig == c("figure1F")) %>%
  ### add log10
  dplyr::mutate(log10_value = log10(value_raw + prior_to_use))

plot_sig_only <- "no"

### To name output files
contrast_name <- "Fig1F"

num_columns_for_plot <- 1

 phen_to_plot <- data_to_use %>%
   dplyr::mutate(TimePoint = factor(TimePoint, c("V2w2L", #DNA/LION
                                                 "V2w2D", #DNA(IM)
                                                 "V2w2EP",#DNA(EP)
                                                 "V2w2DP" #DNA(EP)+protein
                                                 )))
  #### calculate MannWhitney between groups(i.e. TimePoint)
curr_test <- compare_means(log10_value ~ TimePoint, data = phen_to_plot, method = "wilcox.test")
list_of_stats[[paste0(contrast_name)]] <- curr_test

  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$log10_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$log10_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(curr_test),
                          length(which(curr_test[,"p"] < 0.05))
  )
  maxy_y_to_plot <- max_y*(1+0.05*(num_contrasts+1))
  
  color_df <- data.frame("TimePoint"=levels(phen_to_plot$TimePoint),

                         "color"=c("darkred","grey50","black","black"),
                         "shape"=c(16,16,2,6))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = log10_value)) + 
        
    scale_color_manual(values=color_df$color) +
    scale_shape_manual(values=color_df$shape) +
    geom_beeswarm(aes(color = TimePoint, shape = TimePoint), size = 3, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    ## add line with mean
    stat_summary(aes(group = 1, y = log10_value), fun=mean, geom="crossbar", colour="grey30", linetype = "11", size = 0.3)+
    labs(x = "group")+
  theme_bw()+
        theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"))
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(curr_test[l,], label = "p.format",size = 4,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- curr_test[which(curr_test[,"p"]  < 0.05),] 
        
    for(l in 1:nrow(sig_pvals)){
         p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p.format",size = 4,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 plot(p)

Fig 1G

#### this data is already log10
##### select data
data_to_use <- list_of_datasets[["Figure1G"]] %>%
  ### pivot_wider to plot
  tidyr::pivot_wider(names_from = "Antibody", values_from = "value_raw")   %>%
### add treatment group
  dplyr::left_join(list_of_datasets[["Figure1BDEF"]] %>%
                     dplyr::filter(Fig == c("figure1F")) %>%
                     ### add useful treatment
                     dplyr::mutate(Treatment = ifelse(grepl("L",TimePoint),"DNA_LION",
                                                      ifelse(endsWith(TimePoint,"2D"),"DNA","non"))) %>%
                     dplyr::select(MacaqueID,Treatment) %>%
                     dplyr::filter(Treatment != "non") %>%
                     unique(),
                   by = "MacaqueID"
                     ) 

   



### To name output files
contrast_name <- "Fig1G"

num_columns_for_plot <- 1

df_for_plot <- data_to_use %>%
  ## remove unneeded columns
  dplyr::select(-c(Fig,Treatment)) %>%
  column_to_rownames("MacaqueID")

### annotation for columns
annotation_col1 <- data.frame("Antibody_time"=colnames(df_for_plot)) %>%
  ## break up Antibody and timepoint
   dplyr::mutate(Antibody = str_before_first(Antibody_time,"_"),
                 Antibody = ifelse(Antibody == "D614G",Antibody_time,Antibody), ## makes unique colnames
                Time = str_after_first(Antibody_time,"_")) 
# set colnames for df_for_plot
colnames(df_for_plot) <- annotation_col1$Antibody

# trim col annotation
annotation_col <- annotation_col1 %>% 
  dplyr::select(-Antibody_time) %>%
  column_to_rownames("Antibody")

### annotation for rows
annotation_row <- data_to_use %>%
  dplyr::select(c(MacaqueID,Treatment)) %>%
  column_to_rownames("MacaqueID")
  
### color scale
mycols <- colorRampPalette(c("#EBEBEB","#0099D5"))(1000)
breakscale <- seq(0,5, length=1000)


p <- pheatmap(df_for_plot, scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = "black",
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,
              annotation_row = annotation_row,
              display_numbers = round(df_for_plot,1)
              #  annotation_colors = annotation_colors
)
p

Figure 2

Fig 2A

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure2ABDE"]] %>%
  dplyr::filter(Fig %in% c("Fig2A top panel","Fig2 A bottom  panel")) %>%
  ### values are already log10
  dplyr::mutate(log10_value = value_raw,
                group = ifelse(grepl("top",Fig),"nofill",
                                ifelse(grepl(paste(c("V3","V4"),collapse = '|'),TimePoint),"fill","nofill"))) %>%
   ## make timepoint for plotting
   dplyr::mutate(TimePoint = factor(TimePoint,
                                    unique(dplyr::filter(list_of_datasets[["Figure2ABDE"]],
                                                         Fig %in% c("Fig2A top panel","Fig2 A bottom  panel"))$TimePoint))) 





colors_df <- data.frame("panel"=c("Fig2A top panel","Fig2 A bottom  panel"),
                        "color"=c("darkred","grey20"))
fill_df <- data.frame("group"=c("nofill","fill"),
                      "fill"=c("white","#FDC5BC")) %>%
  column_to_rownames("group")

y_limits <-c(0,3)
y_breaks <- c(seq(0,0.9,0.3),seq(1,3,1))
x_labels <- levels(data_to_use$TimePoint) %>%
                       str_replace("w",", w") %>%
                       str_remove_all("k")
x_limits <- levels(data_to_use$TimePoint) 

### To name output files
contrast_name <- "Fig2A"

num_columns_for_plot <- 1

list_of_plots <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i)                    
  curr_color <- colors_df$color[which(colors_df$panel == i)]


 
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = log10_value)) + 
    ## addboxplot
    geom_boxplot(aes(fill = group),color = curr_color)+
        geom_beeswarm(color = curr_color, corral.width = 0.8)+
    scale_fill_manual(values = c("nofill"=fill_df["nofill","fill"],"fill"=fill_df["fill","fill"]))+
    scale_x_discrete(limits = factor(x_limits),
                     labels = x_labels)+
    scale_y_continuous(limits = y_limits,
                       breaks = y_breaks)+
    labs(title = i)+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_blank(), 
            axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

 
      # plot(p)

    list_of_plots[[i]] <- p   
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig 2B

##### set prior for log2 (0 for MSD or Fig 1/2, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["Figure2ABDE"]] %>%
  dplyr::filter(Fig == c("Fig2B")) %>%
  ### values are already log10
  dplyr::mutate(log10_value = value_raw,
                group = ifelse(grepl("top",Fig),"nofill","fill"),
                ### to add spaces in x axis
                TimePoint1 = TimePoint,
                TimePoint2 = factor(TimePoint, unique(dplyr::filter(list_of_datasets[["Figure2ABDE"]],Fig == c("Fig2B"))$TimePoint)),
                TimePoint = factor(ifelse(TimePoint1 == "V2_wk2",as.numeric(TimePoint2)+2,
                                   ifelse(TimePoint1 == "V2_w10",as.numeric(TimePoint2)+3,
                                          ifelse(TimePoint1 == "V2_w42",as.numeric(TimePoint2)+5,
                                                 ifelse(TimePoint1 == "V2_w78",as.numeric(TimePoint2)+8,
                                                        ifelse(TimePoint1 == "V2_w90",as.numeric(TimePoint2)+11,
                                                               as.numeric(TimePoint2)+13))))),seq(0,19,1))
  )

colors_df <- data.frame("panel"=c("Fig2B"),
                        "color"=c("darkblue"))
fill_df <- data.frame("group"=c("nofill","fill"),
                      "fill"=c("white","#B4C0FF")) %>%
  column_to_rownames("group")

y_limits <-c(0,0.5)
y_breaks <- c(seq(0,0.5,0.1))

### breaks are contigous points
x_limits <- seq(0,19,1)
###
x_labels_df <- data.frame("TimePoint"=seq(0,19,1)) %>%
  dplyr::mutate(TimePoint = factor(TimePoint,seq(0,19,1))) %>%
  ## add in actual labels
  dplyr::left_join(unique(data_to_use[,c("TimePoint1","TimePoint")]), by = "TimePoint") 
x_labels <- x_labels_df$TimePoint1 %>%
                       str_after_first("_") 
x_labels[is.na(x_labels)] <- ""

### To name output files
contrast_name <- "Fig2B"

num_columns_for_plot <- 1

list_of_plots <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i)                    
  curr_color <- colors_df$color[which(colors_df$panel == i)]


 
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = log10_value)) + 
    ## addboxplot
    geom_boxplot(aes(fill = group),color = curr_color)+
        geom_beeswarm(color = curr_color, corral.width = 0.8)+
    scale_fill_manual(values = c("nofill"=fill_df["nofill","fill"],"fill"=fill_df["fill","fill"]))+
    scale_x_discrete(limits = factor(x_limits),
                     labels = x_labels)+
    scale_y_continuous(limits = y_limits,
                       breaks = y_breaks)+
    labs(title = i)+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_blank(), 
            axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

 
      # plot(p)

    list_of_plots[[i]] <- p   
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig 2C

##### select data
data_to_use <- list_of_datasets[["Figure2C"]] %>%
  ### add treatment group
  dplyr::left_join(list_of_datasets[["Figure1BDEF"]] %>%
                     dplyr::filter(Fig == c("figure1F")) %>%
                     ### add useful treatment
                     dplyr::mutate(Treatment = ifelse(grepl("L",TimePoint),"DNA_LION",
                                                      ifelse(endsWith(TimePoint,"2D"),"DNA_IM",
                                                             ifelse(grepl("EP",TimePoint),"DNA_EP","DNA_EPplus")))) %>%
                     dplyr::select(MacaqueID,Treatment) %>%
                     unique(),
                   by = "MacaqueID"
                     ) 
 

### To name output files
contrast_name <- "Fig2C"

num_columns_for_plot <- 1

list_of_plots <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i) %>%
   dplyr::mutate(Treatment = factor(Treatment, c("DNA_LION","DNA_IM","DNA_EP","DNA_EPplus")),
                 CellType = factor(CellType, c("CD8","CD4")))

   #### base plot
  p <- ggplot(phen_to_plot, aes(x = MacaqueID, y = value_raw, fill = CellType)) +
    scale_fill_manual(values = c("black","white"))+
    geom_bar(position="stack", stat="identity", color = "black")+
    facet_wrap(~Treatment, 
               strip.position  = "bottom", ## treatment lables are at the bottom
               scales = "free_x", ## don't force the same x ticks in all treatments
               nrow = 1) +
    scale_y_continuous(limits = c(0,3),
                       breaks = c(0,0.3,0.6,0.9,2,3))+
       theme_bw()+
  theme(strip.placement = "outside",
        strip.background = element_rect(fill = NA, color = "white"),
        panel.spacing = unit(-.01,"cm"),
        panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
        axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
        

  
 
    list_of_plots[[i]] <- p   
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig 2DE

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- c("Fig2D","Fig2E")

##### select data
data_to_use <- list_of_datasets[["Figure2ABDE"]] %>%
  ### filter to only vars of interes
  dplyr::filter(grepl(paste(curr_figure_name, collapse = '|'),Fig)) %>%
  dplyr::mutate(Fig = str_replace_all(Fig, " ","_")) 

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::mutate(log2_value = log2(value_raw + as.double(prior_to_use)),
                value_raw = NULL
                
                ) %>%
  dplyr::rename(variable = Fig)


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("w2","w3","w3"),
  "time2"=c("V4","V4","w2")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))


timepoints_x_order <- c("V4","w2","w3")


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig2DE"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 2

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

list_of_pvals <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){

  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each analyte of interest
for(i in unique(data_to_use$Fig)){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$value_raw, na.rm=TRUE)
  min_y <- min(phen_to_plot$value_raw, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = value_raw)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

### reorder list
list_of_plots_ordered <- list()
list_of_plots_ordered[[1]] <- list_of_plots[[1]]
list_of_plots_ordered[[2]] <- list_of_plots[[3]]
list_of_plots_ordered[[3]] <- list_of_plots[[2]]
list_of_plots_ordered[[4]] <- list_of_plots[[4]]

do.call("grid.arrange",c(grobs = list_of_plots_ordered, ncol=num_columns_for_plot))

Fig 2F

##### select data
data_to_use <- list_of_datasets[["Figure2F"]] %>%
  ### calculate averages
  group_by(Fig,Markers,TimePoint) %>%
  dplyr::summarise(mean = mean(value_raw, na.rm = TRUE))


### To name output files
contrast_name <- "Fig2F"

num_columns_for_plot <- 1

list_of_plots <- list()
## for each analyte of interest
for(i in rev(sort(unique(data_to_use$Fig)))){
 phen_to_plot <- data_to_use %>%
    dplyr::filter(Fig == i) %>%
   dplyr::mutate(Markers = factor(Markers, c(
     "EOMES+ T-bet-",
     "EOMES- T-bet-",
     "EOMES+ T-bet+",
     "EOMES- T-bet+")),
                 TimePoint = factor(TimePoint, c("V4","V4w2","V4w3")))

   #### base plot
  p <- ggplot(phen_to_plot, aes(x = TimePoint, y = mean, fill = Markers)) +
    scale_fill_manual(values = c("#919191","#A9A9A9","#00849D","#7AE4EF"))+
    geom_bar(position="stack", stat="identity", color = "black")+
  
       theme_bw()+
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"))
        

  
    list_of_plots[[i]] <- p   
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Figure 3

Fig3A

MSD for DNA/LION Vaccination V1 D1+4hr & D1+24hr

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only V1 D1, D1_1hr, D2
  dplyr::filter(vaccine == "DNA_LION" &
                  vax_dose == "V1" &
                  Day %in% c("D1","D1_4h","D2")
                  ) 


 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_last(sampleName,"__"),
                sampleName = NULL
                ) 

# id_var_of_interest <- "sampleName"
# var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
# ## designate covariates with Day related first, then id
# covariates=c("Day","MacaqueID")


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("V1_D1_4h","V1_D2"),
  "time2"=c("V1_D1","V1_D1")
)

## names for panels (one per row of TimePoints_to_compare)
panel_names <- c("+4hr vs D1 (top)","+24hr vs D1 (bottom)")

## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig3A_MSD_V1"

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

NOTE some p values equal to zero, set to 10^{-18} in stats table for plotting

#### first updata the stats to have 1e-18 for zeros
p_value_df2 <- p_value_df2 %>%
  dplyr::mutate(`V1_D1_4h-V1_D1_pval` = ifelse(`V1_D1_4h-V1_D1_pval` == 0,
                                               as.double(params$prior_for_volcano_p),
                                               `V1_D1_4h-V1_D1_pval`),
                `V1_D2-V1_D1_pval` = ifelse(`V1_D2-V1_D1_pval` == 0,
                                               as.double(params$prior_for_volcano_p),
                                               `V1_D2-V1_D1_pval`))
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

for(i in seq_along(curr_contrasts)){
  c <- curr_contrasts[i]
  label.col <- "rowname"
  sig.col <- paste0(c,raw_or_adj)
  lfc.col <-paste0(c,"_log2FC")
  columns_of_interest <- c(label.col,lfc.col,sig.col)
  
  res.sub <- p_value_df2 %>% dplyr::select(one_of(columns_of_interest)) %>% dplyr::filter(!is.na(!!rlang::sym(lfc.col))) 
  
  change_lfc_name <- "log2FC"
  if(raw_or_adj == "_pval"){
    change_sig_name <- "raw-p"
  }else{
    change_sig_name <- "FDR"
  }
  
  colnames(res.sub) <- c(label.col,change_lfc_name,change_sig_name)
  
  cat(paste0("#### ",panel_names[i],"\n"))
   #Select top genes by logFC or Significance
  no_genes_to_label <- 30
  value_to_sort_the_output_dataset <- "p-value"
  if (value_to_sort_the_output_dataset=="fold-change") {
    res.sub %>% dplyr::arrange(desc(abs(!!rlang::sym(change_lfc_name)))) -> res.sub
  } else if (value_to_sort_the_output_dataset=="p-value") {
    res.sub %>% dplyr::arrange(!!rlang::sym(change_sig_name)) -> res.sub
  }
  genes_to_label <- as.character(res.sub[1:no_genes_to_label,label.col])
  
  additional_labels <- ""
  additional_labels <- unlist(str_split(additional_labels,","))
  filter <- additional_labels %in% res.sub[,label.col]
  additional_labels <- additional_labels[filter]
  missing_labels <- additional_labels[!filter]
  
  if(!is.na(missing_labels)){
    cat("Could not find:\n")
    print(missing_labels)
  }
  use_only_addition_labels <- FALSE
  if(use_only_addition_labels){
    genes_to_label <- additional_labels
  }else{
    genes_to_label <- unique(append(genes_to_label, additional_labels))
  }
  
  
  pCutoff  = 0.05
  FCcutoff = 1.0
  
  significant=as.vector(table(abs( res.sub[,change_lfc_name] ) > FCcutoff &
                                res.sub[,change_sig_name]   < pCutoff))[2]
  
  # fix pvalue == 0
  shapeCustom <- rep(19,nrow(res.sub))
  maxy <-  max(-log10(res.sub[[change_sig_name]]), na.rm=TRUE)
  if(0 > 0){
    maxy <- 0
  }
  
  #cat(paste0("Maxy: ",maxy,"\n"))
  if(maxy == Inf){
    # Sometimes, pvalues == 0
    keep <- res.sub[[change_sig_name]] > 0
    res.sub[[change_sig_name]][!keep] <- min(res.sub[[change_sig_name]][keep])
    shapeCustom[!keep] <- 17
    
    maxy <- -log10(min(res.sub[[change_sig_name]][keep]))
    cat("Some p-values equal zero. Adjusting y-limits.\n")
    cat(paste0("Maxy adjusted: ",maxy,"\n"))
    
  }
  
  # By default, nothing will be greater than maxy. User can set this value lower
  keep <- -log10(res.sub[[change_sig_name]]) <= maxy
  res.sub[[change_sig_name]][!keep] <- maxy
  shapeCustom[!keep] <- 17
  
  names(shapeCustom)<- rep("Exact",length(shapeCustom))
  names(shapeCustom)[shapeCustom == 17] <- "Adjusted"
  
  #Remove if nothin' doin'
  if(all(shapeCustom == 19)){
    shapeCustom <- NULL
  }
  
  maxy <- ceiling(maxy)
  
  use_custom_xlab <- FALSE
  if(use_custom_xlab){
    xlab <- gsub("_"," ",change_lfc_name)
  }else{
    xlab <- bquote(~Log[2]~ "fold change")
  }
  
  xlim_additional <- 0
  ylim_additional <- 0
  axisLabSize <- 12
  legendSize <- 10
  labSize <- 3
  
  imageWidth = 3000
  imageHeight = 3000
  dpi = 300
  
  ####### enhancedvolcano became unusable 9/28/23
  ### add column for signif
  res.sub <- res.sub %>%
    mutate(Significant = ifelse(.data[[change_sig_name]] < pCutoff,
                                ifelse(abs(.data[[change_lfc_name]]) > FCcutoff, "p-value and log2FC","p-value"),
                                ifelse(abs(.data[[change_lfc_name]]) > FCcutoff, "log2FC","NS")),
           Sig_f = factor(Significant,c("NS","log2FC","p-value","p-value and log2FC")))
  
  color_df <- data.frame("category"=c("NS","log2FC","p-value","p-value and log2FC"),
                         "colors"=c("grey60", "forestgreen", "royalblue","red"))
  colors_to_use <- color_df$colors[which(color_df$category %in% unique(res.sub$Sig_f))]
  
p <- ggplot(res.sub,aes(.data[[change_lfc_name]],-log10(.data[[change_sig_name]]),color=Sig_f, fill=Sig_f, label=.data[[label.col]])) +
  geom_point(size = 1, shape = 21, alpha = 0.8) + 
  theme_bw()  +
  geom_text_repel(max.overlaps = 20, size = labSize)+
  geom_hline(yintercept = -log10(pCutoff),linetype="dashed") +
  geom_vline(xintercept = c(-FCcutoff,FCcutoff),linetype="dashed") +
  #scale_x_continuous(limits = c(-15,15))+
  scale_color_manual(values=colors_to_use)+
  scale_fill_manual(values=colors_to_use)+
  theme(legend.position="top", axis.title = element_text(size=axisLabSize), axis.text = element_text(size=axisLabSize), 
        legend.title = element_blank(), legend.text = element_text(size=legendSize),
        
        panel.border = element_blank(), axis.line = element_line(colour = "black", size = 0.5),
        axis.ticks = element_line(size = 0.5), axis.ticks.length = unit(.25,"cm"),
        plot.title = element_text(size = axisLabSize, face = "bold"),
        plot.caption = element_text(size = legendSize)) +
  scale_size(guide = 'none')+ ### remove the size from legend
  guides(color = guide_legend(override.aes = list(size=1)),
         fill = guide_legend(override.aes = aes(label = "")))+ 
  labs(title = paste0("(Significant=",length(which(res.sub$Significant == "p-value and log2FC")),")"),
              caption = paste0("total = ", dim(res.sub)[1], " variables"))
  



  print(p)
  
  
 cat("\n")
cat(' \n \n') ### this is the key!

}

+4hr vs D1 (top)

+24hr vs D1 (bottom)

 cat("\n")

cat(' \n \n') ### this is the key!

Fig 3B

MSD for DNA/LION Vaccination V4 x D2,D3,D4,D8,D15,D22 vs D1

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only V1 D1, D1_1hr, D2
  dplyr::filter(vaccine == "DNA_LION" &
                  vax_dose == "V4" &
                  Day %in% c("D1","D2","D3","D4","D8","D15","D22")
                  ) 


 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_last(sampleName,"__"),
                sampleName = NULL
                ) 

# id_var_of_interest <- "sampleName"
# var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
# ## designate covariates with Day related first, then id
# covariates=c("Day","MacaqueID")


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("V4_D2","V4_D3","V4_D4","V4_D8","V4_D15","V4_D22"),
  "time2"=c("V4_D1","V4_D1","V4_D1","V4_D1","V4_D1","V4_D1")
)


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig3B_MSD_V4"

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

### read in degs results
degs_df <- list_of_stats[["Fig3B_MSD_V4"]]


degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(str_remove_all(colnames(.),"-V4_D1_log2FC"),"V4_")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D2","D3","D4","D8","D15","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D2="grey20",D3="grey40",D4="grey60",D8="grey80",D15="grey90",D22="grey100"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"))

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(timepoint_factor_order,"vsD1_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort
curr_ann_row <- dplyr::arrange(annotation_row[rownames(degs_df_for_hm),],  annotation_row[rownames(degs_df_for_hm),rev(colnames(annotation_row))])

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 3C

MSD for DNA/LION Vaccination V4 D2vsD1, Spearman association among analyte response

prior_to_use <- 0
##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only V1 D1, D1_1hr, D2
  dplyr::filter(vaccine == "DNA_LION" &
                  vax_dose == "V4" &
                  Day %in% c("D1","D2")
                  ) %>%
     ### simplify sampleName
   dplyr::mutate(sampleName = str_remove_all(sampleName,"DNA_LION__V4_")) 




########## generate "dataset" format - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_to_use %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log2_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


cytokine_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_to_use %>%
  ### remove data columns
  dplyr::select(-all_of(cytokine_columns))

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D2-D1")

contrast_name <- "Fig3C_MSD_V4_D2vsD1"
contrast <- "D2-D1$"
contrast_simp <- "DNA_LION_V4_D2vsD1"

 df_original <- diff_counts


  sample_of_interest <- grep(contrast, colnames(df_original), value=TRUE)

  df1 <- df_original %>% select(Gene,one_of(sample_of_interest))
  row.names(df1) <- df1$Gene
  df1$Gene <- NULL

  df <- as.data.frame(t(as.matrix(df1)))

  # Remove
  #remove bad columns
  # filter <- apply(df, 2, function(z) all(is.na(z))) #remove if all NAs
  # df <- df[,!filter]
  filterNA <- apply(df, 2, function(z) any(is.na(z))) #remove if any NAs - better to remove one assay than half the samples
  removedNA <- colnames(df)[which(filterNA)]
  df <- df[,!filterNA]
  removed0 <- ""
  if(length(which(colSums(df)==0)) >0){
    filter0 <- which(colSums(df)==0)
    removed0 <- colnames(df)[filter0]
    df <- df[,-filter0] #remove if all zeros
  }

  ## remove columns with all identical values across the samples
  nonIdentical_colnames <- df %>% dplyr::select(where(~n_distinct(.) > 1)) %>% colnames()
  removedIdentical <- setdiff(colnames(df),nonIdentical_colnames)
  df <- df %>% dplyr::select(where(~n_distinct(.) > 1))
  
  M <- cor(df, method = "spearman")
  p.mat <- cor.mtest(df)$p

  ## check if these are all not sig
  p.mat.adj <- matrix(p.adjust(c(p.mat)), nrow=nrow(p.mat), byrow=FALSE)
  colnames(p.mat.adj)=colnames(df)
  rownames(p.mat.adj)=colnames(df)

  col <- diverging_hcl(n=200, palette="Blue-Red")


  cat(paste0("Number of samples in Corrplot: ",nrow(df)),"\n")

Number of samples in Corrplot: 8

  cat("\n")

  cat(paste0("Number of Analytes in Corrplot: ",ncol(df)),"\n")

Number of Analytes in Corrplot: 41

  cat("\n")

  cat(paste0("Analytes removed for NA: ", removedNA),"\n")

Analytes removed for NA: IL.8

  cat("\n")

  cat(paste0("Analytes removed for zeros: ", removed0),"\n")

Analytes removed for zeros:

  cat("\n")

  cat(paste0("Analytes removed for all identical values: ", removedIdentical),"\n")

Analytes removed for all identical values:

  cat("\n")

  p_val_levels <- c(0.05, 0.01)
  p_val_names <- c("praw","padj")


  # for(i in seq_along(p_val_levels)){
  #   for(j in seq_along(p_val_names)){
  
  i <- 1
  j <- 1
      if(p_val_names[j] == "praw"){
        p.mat_to_use <- p.mat
      }else{
        p.mat_to_use <- p.mat.adj
      }

corrplot(M, method = "ellipse", col = col,
                     type = "upper", order = "hclust", number.cex = 0, #number.font=1,
                     addCoef.col = NULL, # Add coefficient of correlation
                     tl.col = "black", tl.srt = 90, tl.cex=1, cl.cex =1.2,# Text label color and rotation
                     bg = "white", #background color
                     title = paste0(p_val_names[j]," < ",p_val_levels[i]), mar=c(0,0,1,0),
                     # Combine with significance
                     p.mat = p.mat_to_use,
                     sig.level = p_val_levels[i], insig = "blank",
                     # hide correlation coefficient on the principal diagonal
                     diag = FALSE)

Supp Figure S1

Fig S1A

MSD for DNA/LION Vaccination V1, V2, V3, V4 x D1+4h, D2, D4, D8

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only V1 D1, D1_1hr, D2
  dplyr::filter(vaccine == "DNA_LION" &
                  vax_dose %in% c("V1","V2","V3","V4") &
                  Day %in% c("D1","D1_4h","D2","D4","D8")
                  ) 


 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_last(sampleName,"__"),
                sampleName = NULL
                ) 

# id_var_of_interest <- "sampleName"
# var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
# ## designate covariates with Day related first, then id
# covariates=c("Day","MacaqueID")


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("V1_D1_4h","V1_D2","V1_D4","V1_D8",
            "V2_D1_4h","V2_D2","V2_D4","V2_D8",
            "V3_D1_4h","V3_D2","V3_D4","V3_D8",
            "V4_D2","V4_D4","V4_D8"),
  "time2"=c("V1_D1","V1_D1","V1_D1","V1_D1",
            "V2_D1","V2_D1","V2_D1","V2_D1",
            "V3_D1","V3_D1","V3_D1","V3_D1",
            "V4_D1","V4_D1","V4_D1")
)


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "FigS1A_MSD"

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

### read in degs results
degs_df <- list_of_stats[["FigS1A_MSD"]]


degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(str_remove_all(str_remove_all(str_remove_all(
    str_remove_all(colnames(.),"D1_log2FC"),
    "-V1_"),"-V2_"),"-V3_"),"-V4_")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()

#### add in gap for missing
degs_df_for_hm[,"V4_D1_4h"] <- NA


timepoint_factor_order <- c("D1_4h","D2","D4","D8")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Vaccine = c(V1="#038a7a",V2="#78038f",V3="goldenrod1",V4="#e615a0"),
                           Day = c(D1_4h="grey20",D2="grey40",D4="grey60",D8="grey80"))



######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>% 
 dplyr::mutate(
   Day=factor(str_after_first(contrast,"_"),timepoint_factor_order),
   Vaccine = factor(str_before_first(contrast,"_"),c("V1","V2","V3","V4"))) %>%
  dplyr::arrange(Vaccine,Day) %>%
  column_to_rownames("contrast")







############## no row annotation
gaps_col_to_use <- c(4,8,12)

newOrder_row <- read.table(file.path(outputTABLES_dir,paste0(contrast_name,"_rowlist_ordered_hm_ForSpin.txt")), header=T)
newOrder_row_v <-as.vector(newOrder_row[,3])

  

p <- pheatmap(degs_df_for_hm[,rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
               cluster_rows=dendextend::rotate(myMap$tree_row, newOrder_row_v),
                gaps_col = gaps_col_to_use, 
              border_color = NA,
                #cluster_rows = FALSE, 
              cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              #annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig S1B

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only DNA_LION
  ### all timepoints for each dose
  dplyr::filter(vaccine == "DNA_LION" &
                  vax_dose %in% c("V1","V2","V3","V4") 
                  ) 





df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out day & macaque ID
  dplyr::mutate(vax_dose_day = factor(str_after_last(sampleName,"__"),
                                  c("V1_D1","V1_D1_4h","V1_D2","V1_D4","V1_D8",
                                    "V2_D1","V2_D1_4h","V2_D2","V2_D4","V2_D8",
                                    "V3_D1","V3_D1_4h","V3_D2","V3_D4","V3_D8",
                                    "V4_D1","V4_D2","V4_D3","V4_D4","V4_D8",
                                    "V4_D15","V4_D22")),
                ### to add spaces
                vax_dose_day2 = ifelse(grepl("V1",vax_dose_day),as.numeric(vax_dose_day),
                                       ifelse(grepl("V2",vax_dose_day),as.numeric(vax_dose_day)+1,
                                              ifelse(grepl("V3",vax_dose_day),as.numeric(vax_dose_day)+2,
                                              as.numeric(vax_dose_day)+3))),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL,
                ### add cat for spacing between vax doses
                ) 


selected_analytes <- c("IL.15","CXCL10.IP.10",
                       "IFN.g","CCL4.MIP.1b",
                       "IL.1Ra","IL.6",
                       "FLT3L","CXCL13",
                       "CXCL11.ITAC","CCL2.MCP.1")

### To name output files
contrast_name <- "FigS1B_MSD"

num_columns_for_plot <- 2

list_of_plots_mean <- list()
list_of_plots_median <- list()
## for each analyte of interest
for(i in selected_analytes){
 phen_to_plot <- df_long %>%
    dplyr::filter(variable == i)
  
 
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = vax_dose_day2, y = raw_value, group = MacaqueID)) + 
    geom_line(color = "grey60")+
    ## adds spaces in x axis
    scale_x_continuous(breaks = unique(df_long$vax_dose_day2),
                     #labels = str_replace(levels(phen_to_plot$vax_dose_day),"_","\n")
                     labels = str_replace(levels(phen_to_plot$vax_dose_day),"_",", "))+
    labs(title = i)+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_blank(), 
            axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
  
  p_mean <- p +
    stat_summary(aes(group = 1, y = raw_value), fun=mean, geom="line", colour="red")
  p_median <- p +
    stat_summary(aes(group = 1, y = raw_value), fun=median, geom="line", colour="red")

      # plot(p)
  list_of_plots_mean[[i]] <- p_mean
  list_of_plots_median[[i]] <- p_median  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

  cat("#### ",paste0("Mean"),"\n")

Mean

  cat('\n')

do.call("grid.arrange",c(list_of_plots_mean, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

    cat("#### ",paste0("Median"),"\n")

Median

  cat('\n')

do.call("grid.arrange",c(list_of_plots_median, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig S1C

MSD for DNA/LION Vaccination V4 x D2, D4, D8, D15 & D22 vs D1

### stats are the same as for Fig3B_MSD_V4
list_of_stats[["FigS1C_MSD_V4"]] <- list_of_stats[["Fig3B_MSD_V4"]]

degs_df <- list_of_stats[["FigS1C_MSD_V4"]]


degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(str_remove_all(colnames(.),"-V4_D1_log2FC"),"V4_")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("V4_D2","V4_D3","V4_D4","V4_D8","V4_D15","V4_D22"),
  "time2"=c("V4_D1","V4_D1","V4_D1","V4_D1","V4_D1","V4_D1")
)

timepoint_factor_order <- c("D2","D3","D4","D8","D15","D22")



############## use row annotation with sig categories as alluvial df

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)


######### make long dataframe to plot
combo_df_long <- annotation_row %>%
  ### remove rows with all NA
  filter(if_any(everything(), ~ !is.na(.))) %>%
  rownames_to_column("Gene") %>%
  tidyr::pivot_longer(!Gene, names_to = "timepoint", values_to = "dir") %>%
  dplyr::mutate(timepoint = factor(str_remove_all(timepoint,"_dir"),c("D2vsD1","D3vsD1","D4vsD1","D8vsD1","D15vsD1","D22vsD1")),
                Gene = factor(Gene))


 ###### set colors
   color_df <- data.frame("group"=unique(levels(combo_df_long$dir)),
                        "color"=c("red","lightpink","grey85","skyblue","blue"))
 
  curr_color_df <- color_df[which(color_df$group %in% unique(combo_df_long$dir)),]

 ####### plot
      gg <- ggplot(combo_df_long,
               aes(x = timepoint, stratum = dir, alluvium = Gene))+
        stat_alluvium(aes(fill = dir), width = .65, alpha = 1/2,) +
        stat_stratum(aes(stratum = dir, fill = dir), width = .65) +
        stat_alluvium(geom = "text", aes(label = Gene), size = 6.3) +
        scale_fill_manual(values = curr_color_df$color)+
    theme_bw()+
      theme(axis.text=element_text(size=20),legend.text=element_text(size=20),
        axis.title=element_text(size=20,face="bold"))
    #ggtitle(paste0(tissue_to_use))

   plot(gg)

    cat("\n")

  cat(' \n \n') ### this is the key!

Supp Figure S2

Fig S2B

MSD for DNA/LION Vaccination & DNA Vaccination V1, V2 x D2, D4, D8

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only V1 D1, D1_1hr, D2
  dplyr::filter(vaccine %in% c("DNA_LION","DNA") &
                  vax_dose %in% c("V1","V2") &
                  Day %in% c("D1","D2","D4","D8")
                  ) 


 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 

# id_var_of_interest <- "sampleName"
# var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
# ## designate covariates with Day related first, then id
# covariates=c("Day","MacaqueID")


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("DNA_LION__V1_D2","DNA_LION__V1_D4","DNA_LION__V1_D8",
            "DNA_LION__V2_D2","DNA_LION__V2_D4","DNA_LION__V2_D8",
            "DNA__V1_D2","DNA__V1_D4","DNA__V1_D8",
            "DNA__V2_D2","DNA__V2_D4","DNA__V2_D8"),
  "time2"=c("DNA_LION__V1_D1","DNA_LION__V1_D1","DNA_LION__V1_D1",
            "DNA_LION__V2_D1","DNA_LION__V2_D1","DNA_LION__V2_D1",
            "DNA__V1_D1","DNA__V1_D1","DNA__V1_D1",
            "DNA__V2_D1","DNA__V2_D1","DNA__V2_D1")
)


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "FigS2B_MSD"

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

### read in degs results
degs_df <- list_of_stats[["FigS2B_MSD"]]


degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(
    str_remove_all(
      str_remove_all(
      str_remove_all(
    str_remove_all(colnames(.),"V1_D1_log2FC"),
    "V2_D1_log2FC"),"-DNA_LION__"),"-DNA__")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()




timepoint_factor_order <- c("D2","D4","D8")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Vaccine = c(DNA_LION="darkgoldenrod",DNA="darkolivegreen"),
                           Dose = c(V1="#038a7a",V2="#78038f"),
                           Day = c(D2="grey40",D4="grey60",D8="grey80"))



######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>% 
 dplyr::mutate(
   Vaccine = factor(str_before_first(contrast,"__"),c("DNA_LION","DNA")),
   Dose = factor(str_before_first(str_after_first(contrast,"__"),"_"),c("V1","V2")),
   Day=factor(str_after_last(contrast,"_"),timepoint_factor_order)) %>%
  dplyr::arrange(Vaccine,Dose,Day) %>%
  column_to_rownames("contrast")



############## no row annotation
gaps_col_to_use <- c(3,6,9)

newOrder_row <- read.table(file.path(outputTABLES_dir,paste0(contrast_name,"_rowlist_ordered_hm_ForSpin.txt")), header=T)
newOrder_row_v <-as.vector(newOrder_row[,3])

  

p <- pheatmap(degs_df_for_hm[,rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
               cluster_rows=dendextend::rotate(myMap$tree_row, newOrder_row_v),
                gaps_col = gaps_col_to_use, 
              border_color = NA,
                #cluster_rows = FALSE, 
              cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              #annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig S2C

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 0

##### select data
data_to_use <- list_of_datasets[["MSD"]] %>%
  ### filter to only DNA_LION
  ### only V1, V2; D1,D2,D4,D8
  dplyr::filter(vaccine %in% c("DNA_LION","DNA") &
                  vax_dose %in% c("V1","V2") &
                  Day %in% c("D1","D2","D4","D8")
                  ) 



df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,vaccine,vax_dose,Day)) %>%
  ### make longer 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out day & macaque ID
  dplyr::mutate(vax_dose_day = factor(str_replace_all(str_after_first(sampleName,"__"),"DNA_LION","DNALION"),
                                  c("DNALION__V1_D1","DNALION__V1_D2","DNALION__V1_D4","DNALION__V1_D8",
                                    "DNALION__V2_D1","DNALION__V2_D2","DNALION__V2_D4","DNALION__V2_D8",
                                    "DNA__V1_D1","DNA__V1_D2","DNA__V1_D4","DNA__V1_D8",
                                    "DNA__V2_D1","DNA__V2_D2","DNA__V2_D4","DNA__V2_D8")),
                ### to add spaces
                vax_dose_day2 = ifelse(grepl("DNALION__V1",vax_dose_day),as.numeric(vax_dose_day),
                                       ifelse(grepl("DNALION__V2",vax_dose_day),as.numeric(vax_dose_day)+1,
                                              ifelse(grepl("DNA__V1",vax_dose_day),as.numeric(vax_dose_day)+2,
                                              as.numeric(vax_dose_day)+3))),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL,
                ### add cat for spacing between vax doses
                ) 


selected_analytes <- c("CX3CL1.Fractalkine","CXCL11.ITAC","IL.15","IL.7",
                       "IL.1Ra","CXCL10.IP.10","CXCL13","CCL2.MCP.1",
                       "FLT3L","IL.18")

### To name output files
contrast_name <- "FigS2C_MSD"

num_columns_for_plot <- 4

list_of_plots <- list()

## for each analyte of interest
for(i in selected_analytes){
 phen_to_plot <- df_long %>%
    dplyr::filter(variable == i)
  
 
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = vax_dose_day2, y = raw_value, group = MacaqueID)) + 
    geom_line(color = "grey20")+
    geom_point()+
    ## adds spaces in x axis
    scale_x_continuous(breaks = unique(df_long$vax_dose_day2),
                     #labels = str_replace(levels(phen_to_plot$vax_dose_day),"_","\n")
                     labels = str_replace(levels(phen_to_plot$vax_dose_day),"__",", "))+
    labs(title = i)+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_blank(), 
            axis.title.x = element_blank(),
            axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

      # plot(p)
  list_of_plots[[i]] <- p

  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

  cat("\n")

  cat(' \n \n') ### this is the key!

Figure 4

Fig 4C

Broad cell populations vs Time

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig4ABC"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))
  


 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))

timepoints_x_order <- c("D1","D3","D8","D22")


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig4C"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

### order for scatterplots
variables_ordered <- c("HLADRpos_Myeloid","B","T")

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),variables_ordered),
                Day = factor(str_after_first(sampleName,"__"),timepoints_x_order),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 4D

B:T cell ratio in Blood vs LN

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig4D"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))
  


 
### generate dataframe: rows = samples, columns = variables (including timepoint) and group for comparison (= compartment)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_before_first(variable,"__"), ### call timepoint the variable that you want to compare
                variable = paste0(str_after_first(variable,"__"),"__",str_after_first(sampleName,"__")), ### add timepoint to variable
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("LN"),
  "time2"=c("blood")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig4D"

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = str_after_first(variable,"__"),
                Day = factor(str_after_first(sampleName,"__"),c("D1","D3","D8","D22")),
                MacaqueID = str_before_first(sampleName,"__"),
                var_tp = paste0(variable, "__", Day),
                sampleName = NULL) %>%
  dplyr::arrange(Day, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


plot_sig_only <- "no"

 phen_to_plot <- df_long 
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.08*(num_contrasts+1)),
                                  max_y*(1+0.08*(num_contrasts+1))
                                         
                                         )
  
  
  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
        scale_fill_manual(values=c("white","black")) +  

     geom_beeswarm(aes(fill=compartment), color = "black", size = 2, shape=21, alpha=0.8, cex = 3, dodge.width = 0.7)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){


  
  for(l in 1:length(list_of_pvals)){
    currpvals <- list_of_pvals[[l]] %>% 
    dplyr::mutate(x = as.numeric(l),
                  xmin = x-0.2,
                  xmax = x+0.2)
    
     p <- p +
        stat_pvalue_manual(currpvals[1,], label = "p",size = 2.5,
                           y.position = max_y+0.1*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  plot(p)

  cat("\n")

  cat(' \n \n') ### this is the key!

Fig 4E

B-cell folicle IHC

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig4E"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__"))))
  
##### controls are 3 different animals for GCtoNonGC only, not used for stats

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))

timepoints_x_order <- c("D1","D3","D8","D22")


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig4E"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

### variables ordered in plots
variables_ordered <- c("LN__BCF_GC","LN__BCF_noGC","LN__GCtoNonGC_ratio")

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),variables_ordered),
                Day = factor(str_after_first(sampleName,"__"),timepoints_x_order),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Supp Figure S3

CD4 and CD8 T cells

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "FigS3"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))

timepoints_x_order <- c("D1","D3","D8","D22")


## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "FigS3"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 2

### variables ordered in plots
variables_ordered <- c("CD4","CD8")

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),variables_ordered),
                Day = factor(str_after_first(sampleName,"__"),timepoints_x_order),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Figure 5

Myeloid populations

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig5"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig5ABC"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

Fig 5A

varcomp_to_plot <- c("blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos","LN__Pct_of_HLADRpos_HLA_DRpos_Ki67pos")

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                Day = factor(str_after_first(sampleName,"__"),c("D1","D3","D8","D22")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 5B

varcomp_to_plot <- c("blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos","LN__Pct_of_HLADRpos_HLA_DRpos_Ki67pos",
                     "blood__Pct_of_HLADRpos_pDC_Ki67pos","LN__Pct_of_HLADRpos_pDC_Ki67pos",
                     "blood__Pct_of_HLADRpos_cDC_Ki67pos","LN__Pct_of_HLADRpos_cDC_Ki67pos",
                     "blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos","LN__Pct_of_HLADRpos_Classical_Mo_Ki67pos")

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 5C

varcomp_to_plot <- c("blood__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos","LN__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos",
                     "blood__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos","LN__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                     "blood__Pct_of_HLADRpos_DC3_Ki67pos","LN__Pct_of_HLADRpos_DC3_Ki67pos",
                     "blood__Pct_of_HLADRpos_cDC1_Ki67pos","LN__Pct_of_HLADRpos_cDC1_Ki67pos",
                     "blood__Pct_of_HLADRpos_cDC2_Ki67pos","LN__Pct_of_HLADRpos_cDC2_Ki67pos")

num_columns_for_plot <- 5

lineplot

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                Day = factor(str_after_first(sampleName,"__"),c("D1","D3","D8","D22")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

heatmap

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 5D

Spearman correlation between MSD D2vsD1 and Myeloid populations

########## generate "dataset" format - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_to_use %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log2_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


analyte_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_to_use %>%
  ### remove data columns
  dplyr::select(-all_of(analyte_columns))

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D3-D1","D8-D1","D22-D1")

contrast_name <- "Fig5DE_Myeloid"

raw_or_adj <- "raw"

### which dataset levels/FC, trimmed to the tab
curr_ds <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  column_to_rownames("Gene")

###### MSD = D2-D1 only log2FC
curr_df2 <- data.frame(t(list_of_diff_counts[[paste0("Fig3C_MSD_V4_D2vsD1")]])) %>%
  ### make row 1 colnames
  janitor::row_to_names(row_number = 1) %>%
  ### trim to only D2-D1
  rownames_to_column("sampleName") %>%
  dplyr::filter(grepl("D2-D1",sampleName)) %>%
  ### be sure sorted
  dplyr::arrange(sampleName) %>%
  column_to_rownames("sampleName") %>%
  #### then remove columns with all NA
  select_if(~ !all(is.na(.))) %>%
  ### be sure numeric
  mutate_if(is.character, as.numeric)
  

set_names_to_use <- c(paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC"))

corr_tag_name <- "CorrVsMSD"

 ######## calculate correlation
  cor_method <- "spearman"
  
  CorrName <- paste0(contrast_name,"__",corr_tag_name)
    
  #### generate empty dataframe for combo
fullcombo_df_long <- data.frame()


  for(set_name in set_names_to_use){
   curr_df1 <- data.frame(t(curr_ds)) %>%
    
    ### trim to only tp of interest
    rownames_to_column("sampleName") %>%
    dplyr::filter(grepl(str_before_first(set_name,"_"),sampleName)) %>%
    ### be sure sorted
    dplyr::arrange(sampleName) %>%
    column_to_rownames("sampleName") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ## remove rows with all NAs
    filter(if_any(everything(), ~ !is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)
  

    ## generate cor matrix 
  cor_mat <- cor(as.matrix(signif(curr_df1,7)), as.matrix(signif(curr_df2,7)), method = cor_method, use="pa")
 
  ## raw p
  cor_test_mat <- data.frame(matrix(NA, nrow=dim(curr_df1)[2], ncol=dim(curr_df2)[2]))
  rownames(cor_test_mat) <- colnames(curr_df1)
  colnames(cor_test_mat) <- colnames(curr_df2)
  
  for(curr1 in colnames(curr_df1)) {
    for(curr2 in colnames(curr_df2)) {
      cor_test_mat[curr1, curr2] <- 
        cor.test(as.matrix(signif(curr_df1,7)[,curr1]), 
                 as.matrix(signif(curr_df2,7)[,curr2]), 
                 method = cor_method, use="pa")$p.value
    }
  }
  
 ## adj p for multiple comparisons by BH over cytokine assays
  cor_test_adj_mat <- cor_test_mat #duplicate dataframe to replace values
  for(curr_assay in colnames(curr_df1)) {
    cor_test_adj_mat[curr_assay,] <- p.adjust(cor_test_mat[curr_assay,], method = "BH")
  }


  
    ### write function for converting the cor_mat to long format df
    corr_to_long <- function(mat){
      data.frame(mat) %>%
        rownames_to_column("celltype") %>%
        tidyr::pivot_longer(!celltype, names_to = "assay", values_to = "value") %>%
        dplyr::mutate(celltype_assay = paste0(celltype,"_vs_",assay))
      
    }

    
    
    
    
    #### convert to long
    cor_long <-  corr_to_long(cor_mat)   
    if(raw_or_adj == "raw"){
      p_long <-  corr_to_long(cor_test_mat)     
    }else{
      p_long <-  corr_to_long(cor_test_adj_mat)     
    }
    
    #### combine cor and p long dfs
    curr_df_long <- dplyr::left_join(cor_long,p_long, by = c("celltype","assay","celltype_assay")) %>%
      ### rename values to be useful
      dplyr::rename(corr_coef=value.x,p=value.y) %>%
      ### add column with set_name
      dplyr::mutate(timepoint = paste0(str_replace_all(set_name,"-","vs")))

  
  ### add to fullcombo
  fullcombo_df_long <- rbind(fullcombo_df_long,curr_df_long)
  
  }

### add to stats master list
list_of_stats[[paste0(CorrName)]] <- fullcombo_df_long %>% 
  dplyr::relocate(celltype, assay, corr_coef, p, timepoint)

blood

barbara_order <- c("blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos",
                   "blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_DC3_Ki67pos"
                   )
 

## only for the Intervals
timepoints_to_use <- paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC")
TP_name <- "Intervals"

corr_tag_name <- "CorrVsMSD" 
raw_or_adj <- "raw"
cor_method <- "spearman"
trim_pval_level <- 0.05
all_or_selected <- "selected"
assay_arrangement <- c("CCL19.MIP.3b","CCL24.Eotaxin.2","CCL26.Eotaxin.3","CCL27.CTACK","CXCL13",
                       "FLT3L","CCL13.MCP.4","CXCL12.SDF.1a","CCL11.Eotaxin","CCL22.MDC",
                       
                       #"IL.17B",
                       "IL.9") ### if trimming/ordering assays


### set pos/neg/both
directions_to_use <- c("pos")
assays_to_exclude <- c("")

raw p < 0.05

#### make df for plotting 
df_long1 <- fullcombo_df_long %>%
  ### generate dir column
  dplyr::mutate(dir = ifelse(p < trim_pval_level,
                             ifelse(corr_coef > 0, "pos","neg"),
                             "notSig"),
                ### break out compartment
                compartment = str_before_first(celltype,"__"),
                celltype = str_after_first(celltype,"__"),
                celltype_assay = str_after_first(celltype_assay,"__")) %>%
  ### remove the corr and p vals for pivoting
  dplyr::select(-c(corr_coef,p)) 



###### set up inital dataframe
## trim to only sig
bub_df2 <- df_long1[which(df_long1$dir %in% directions_to_use ),] %>%
  ### also remove any analytes_to_exclude
  dplyr::filter(!assay %in% assays_to_exclude) %>%
  dplyr::mutate(timepoint = factor(str_before_first(timepoint,"_"),str_replace_all(str_before_first(timepoints_to_use,"_"),"-","vs")),
                dir = factor(dir, directions_to_use),
                ##### add compartment to celltype
                comp_celltype = paste0(compartment,"__",celltype)) 


#### remove timepoint suffix from celltype (only needed for vs Abs/NAbs)
cleaned_celltypes <- bub_df2$comp_celltype
for(tu in timepoints_to_use){
  cleaned_celltypes <- str_remove_all(cleaned_celltypes,paste0("__",str_replace_all(tu,"-","vs")))
}

bub_df2$celltype <- cleaned_celltypes




  #### if selected tab combos, keep cell pops on the x-axis

      
         bub_df_bothordered <- bub_df2 %>%
           #### trim to desired assays
           dplyr::filter(assay %in% assay_arrangement) %>%
      ### remove factor levels for assay
      dplyr::mutate(assay = factor(as.character(assay),assay_arrangement),
                    ### also breakout for celltype
                    celltype_only = str_after_first(celltype,"__"),
                    compartment = factor(compartment, c("blood","LN"))) %>%   
      dplyr::arrange(assay) %>%
      #### trim to barbara_order celltypes
      dplyr::filter(celltype %in% barbara_order) %>%
      ## lock in factor levels for assay (rev for y axis since want it to be reversed)
      dplyr::mutate(assay = factor(assay, rev(unique(assay))),
                    ### now get celltype order
                    celltype = factor(celltype, intersect(barbara_order,unique(celltype)))) %>%
           #### now arrange by timepoint to be able to see bulls eyes
           dplyr::arrange(desc(timepoint))
    
 
      ###### reversed axes!!!
    bba <- 
      ggplot(bub_df_bothordered, aes(x = celltype, y = assay))+
      geom_point(aes(size = timepoint, color = timepoint))+
      #scale_alpha_manual(values = alpha_values)+
      scale_size_manual(values = c(3,6,10))+
      scale_color_manual(values = c("#262626","#7F7F7F","#BFBFBF"))+
      theme_bw()+
      ### make legend symbol bigger & fix color 
      # guides(color = guide_legend(override.aes = list(size = 5)),
      #        size = guide_legend(override.aes = list(color = leg_color)))+
      ## remove gridlines
      theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            ## rotate axis labels
            axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
  
  
    plot(bba)

  cat("\n")

    cat(' \n \n') ### this is the key!

      ### last plot in loop doesn't appear unless add this
  
    
 
 
  cat("\n")

    cat(' \n \n') ### this is the key!

    ###### need to fix why this plots in Rstudio but not renedered

LN

barbara_order <- c("blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos",
                   "blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                   "blood__Pct_of_HLADRpos_DC3_Ki67pos"
                   )
 

## only for the Intervals
timepoints_to_use <- paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC")
TP_name <- "Intervals"

corr_tag_name <- "CorrVsMSD" 
raw_or_adj <- "raw"
cor_method <- "spearman"
trim_pval_level <- 0.05
all_or_selected <- "selected"
assay_arrangement <- c("CCL19.MIP.3b","CCL24.Eotaxin.2","CCL26.Eotaxin.3","CCL27.CTACK","CXCL13",
                       "FLT3L","CCL13.MCP.4","CXCL12.SDF.1a","CCL11.Eotaxin","CCL22.MDC",
                       
                       #"IL.17B",
                       "IL.9") ### if trimming/ordering assays


### set pos/neg/both
directions_to_use <- c("pos")
assays_to_exclude <- c("")

raw p < 0.05

#### make df for plotting 
df_long1 <- fullcombo_df_long %>%
  ### generate dir column
  dplyr::mutate(dir = ifelse(p < trim_pval_level,
                             ifelse(corr_coef > 0, "pos","neg"),
                             "notSig"),
                ### break out compartment
                compartment = str_before_first(celltype,"__"),
                celltype = str_after_first(celltype,"__"),
                celltype_assay = str_after_first(celltype_assay,"__")) %>%
  ### remove the corr and p vals for pivoting
  dplyr::select(-c(corr_coef,p)) 



###### set up inital dataframe
## trim to only sig
bub_df2 <- df_long1[which(df_long1$dir %in% directions_to_use ),] %>%
  ### also remove any analytes_to_exclude
  dplyr::filter(!assay %in% assays_to_exclude) %>%
  dplyr::mutate(timepoint = factor(str_before_first(timepoint,"_"),str_replace_all(str_before_first(timepoints_to_use,"_"),"-","vs")),
                dir = factor(dir, directions_to_use),
                ##### add compartment to celltype
                comp_celltype = paste0(compartment,"__",celltype)) 


#### remove timepoint suffix from celltype (only needed for vs Abs/NAbs)
cleaned_celltypes <- bub_df2$comp_celltype
for(tu in timepoints_to_use){
  cleaned_celltypes <- str_remove_all(cleaned_celltypes,paste0("__",str_replace_all(tu,"-","vs")))
}

bub_df2$celltype <- cleaned_celltypes




  #### if selected tab combos, keep cell pops on the x-axis

      
         bub_df_bothordered <- bub_df2 %>%
           #### trim to desired assays
           dplyr::filter(assay %in% assay_arrangement) %>%
      ### remove factor levels for assay
      dplyr::mutate(assay = factor(as.character(assay),assay_arrangement),
                    ### also breakout for celltype
                    celltype_only = str_after_first(celltype,"__"),
                    compartment = factor(compartment, c("blood","LN"))) %>%   
      dplyr::arrange(assay) %>%
      #### trim to barbara_order celltypes
      dplyr::filter(celltype %in% barbara_order) %>%
      ## lock in factor levels for assay (rev for y axis since want it to be reversed)
      dplyr::mutate(assay = factor(assay, rev(unique(assay))),
                    ### now get celltype order
                    celltype = factor(celltype, intersect(barbara_order,unique(celltype)))) %>%
           #### now arrange by timepoint to be able to see bulls eyes
           dplyr::arrange(desc(timepoint))
    
 
      ###### reversed axes!!!
    bba <- 
      ggplot(bub_df_bothordered, aes(x = celltype, y = assay))+
      geom_point(aes(size = timepoint, color = timepoint))+
      #scale_alpha_manual(values = alpha_values)+
      scale_size_manual(values = c(3,6,10))+
      scale_color_manual(values = c("#262626","#7F7F7F","#BFBFBF"))+
      theme_bw()+
      ### make legend symbol bigger & fix color 
      # guides(color = guide_legend(override.aes = list(size = 5)),
      #        size = guide_legend(override.aes = list(color = leg_color)))+
      ## remove gridlines
      theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            ## rotate axis labels
            axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
  
  
    plot(bba)

  cat("\n")

    cat(' \n \n') ### this is the key!

      ### last plot in loop doesn't appear unless add this
  
    
 
 
  cat("\n")

    cat(' \n \n') ### this is the key!

    ###### need to fix why this plots in Rstudio but not renedered

Fig 5E

Selected scatterplots

selected_pairs_df <- data.frame("var_x"=c("blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos",
                                          "blood__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                                          "blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos",
                                          "blood__Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                                          "LN__Pct_of_HLADRpos_Classical_Mo_Ki67pos",
                                          "LN__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos",
                                          "LN__Pct_of_HLADRpos_DC3_Ki67pos",
                                          "LN__Pct_of_HLADRpos_Intermediate_Mo_Ki67pos"),
                                "Day_x"=c("D3-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D3-D1",
                                             "D8-D1",
                                             "D3-D1"
                                             ),
                                "var_y"=c("IL.9",
                                          "FLT3L",
                                          "CCL27.CTACK",
                                          "CXCL12.SDF.1a",
                                          "G.CSF",
                                          "CXCL12.SDF.1a",
                                          "FLT3L",
                                          "CCL13.MCP.4"),
                                "Day_y"=c(rep("D2-D1",8))
)

num_columns_for_plot <- 2

### pivot curr_ds to separate timepoint from animal
df_long <- curr_ds %>%
  rownames_to_column("comp_var") %>%
  tidyr::pivot_longer(!comp_var, names_to = "sampleName", values_to = "log2FC") %>%
  ### break out MacaqueID and Day
  dplyr::mutate(Day = str_after_first(sampleName,"__"),
                Day = str_replace_all(Day,"-","vs"),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  ### add in the second df
  dplyr::full_join(data.frame(t(curr_df2)) %>%
                     rownames_to_column("comp_var") %>%
                     tidyr::pivot_longer(!comp_var, 
                                         names_to = "sampleName", values_to = "log2FC") %>%
                     ### break out MacaqueID and Day
                     dplyr::mutate(Day = str_after_first(sampleName,"__"),
                                   Day = str_replace_all(Day,"\\.","vs"),
                                   Day = str_replace_all(Day,"-","vs"),
                                   MacaqueID = str_before_first(sampleName,"__"),
                                   sampleName = NULL),
                   by = c("comp_var","Day","MacaqueID","log2FC")
  )



list_of_plots <- list()
## for each selected pair
for(i in 1:nrow(selected_pairs_df)){
  curr_x <- selected_pairs_df$var_x[i]
  curr_Day_x <- str_replace_all(selected_pairs_df$Day_x[i],"-","vs")
  curr_y <- selected_pairs_df$var_y[i]
  curr_Day_y <- str_replace_all(selected_pairs_df$Day_y[i],"-","vs")
  
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var %in% c(curr_x,curr_y)) %>%
   dplyr::filter(Day %in% c(curr_Day_x,curr_Day_y)) %>%
   ### combine var and time
   dplyr::mutate(comp_var_day = paste0(comp_var,"___",Day),
                 comp_var = NULL, Day = NULL) %>%
   tidyr::pivot_wider(names_from = comp_var_day, values_from = log2FC)
  

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = !!rlang::sym(paste0(curr_x,"___",curr_Day_x)), 
                                y = !!rlang::sym(paste0(curr_y,"___",curr_Day_y)))) + 
    geom_point()+
    stat_cor(method = "spearman")+
    xlab(paste0(curr_x,"\n",curr_Day_x))+
    ylab(paste0(curr_y,"\n",curr_Day_y))+
           
    theme_bw()+
    theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title = element_text(size = 10)) ### make smaller to fit
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Supp Figure S4

Fig S4A

varcomp_to_plot <- c("blood__Pct_of_HLADRpos_pDC_Ki67pos","LN__Pct_of_HLADRpos_pDC_Ki67pos",
                     "blood__Pct_of_HLADRpos_cDC_Ki67pos","LN__Pct_of_HLADRpos_cDC_Ki67pos",
                     "blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos","LN__Pct_of_HLADRpos_Classical_Mo_Ki67pos")

#### need to reload Fig5 prep
list_of_stats[["FigS4A"]] <- data.frame("FigS4A"="stats are in the Fig5ABC tab")

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig5"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig5ABC"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                Day = factor(str_after_first(sampleName,"__"),c("D1","D3","D8","D22")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

FigS4B

########## uses data from Figure 5
##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig5"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))



curr_phen_long  <- data_to_use %>%
  # calculate additional populations for generating proportions of 100%
  dplyr::mutate(LN__Pct_of_HLADRpos_cDC_wo_cDC1orcDC2_Ki67pos = LN__Pct_of_HLADRpos_cDC_Ki67pos -
                  (LN__Pct_of_HLADRpos_cDC1_Ki67pos + LN__Pct_of_HLADRpos_cDC2_Ki67pos),
                LN__Pct_of_HLADRpos_cDC_Ki67pos = NULL,
                LN__Pct_of_HLADRpos_Classical_Mo_minusDC3_Ki67pos = LN__Pct_of_HLADRpos_Classical_Mo_Ki67pos - LN__Pct_of_HLADRpos_DC3_Ki67pos,
                LN__Pct_of_HLADRpos_Classical_Mo_Ki67pos = NULL,
                LN__Pct_of_HLADRpos_HLA_DRpos_Ki67pos = NULL,
                
                blood__Pct_of_HLADRpos_cDC_wo_cDC1orcDC2_Ki67pos = blood__Pct_of_HLADRpos_cDC_Ki67pos -
                  (blood__Pct_of_HLADRpos_cDC1_Ki67pos + blood__Pct_of_HLADRpos_cDC2_Ki67pos),
                blood__Pct_of_HLADRpos_cDC_Ki67pos = NULL,
                blood__Pct_of_HLADRpos_Classical_Mo_minusDC3_Ki67pos = blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos - blood__Pct_of_HLADRpos_DC3_Ki67pos,
                blood__Pct_of_HLADRpos_Classical_Mo_Ki67pos = NULL,
                blood__Pct_of_HLADRpos_HLA_DRpos_Ki67pos = NULL) %>%
  ### get average for each day/compartment
  # first make longer
  dplyr::select(-sampleName) %>%
  tidyr::pivot_longer(cols = contains("__"), names_to = "comp_var", values_to = "value") %>%
  dplyr::mutate(compartment = factor(str_before_first(comp_var,"__"), c("blood","LN")),
                variable = str_after_last(comp_var,"__"),
                Day = factor(Day, c("D1","D3","D8","D22"))) %>%

  
  
  ### collect proportion of 100% and means
  # collect total
  group_by(compartment, Day, MacaqueID) %>%
  dplyr::mutate(total_per_comp_time = sum(value)) %>%
  ungroup() %>%
  # calculate prop within each animal
  group_by(compartment, variable, Day, MacaqueID) %>%
  dplyr::mutate(prop = value / total_per_comp_time *100) %>%  
  ungroup() %>%
  # take mean across animals
  group_by(compartment, variable, Day) %>%
  dplyr::summarise(mean = mean(prop)) %>%
  # add comp_time
  dplyr::mutate(comp_time = paste0(compartment,"__",Day),
                variable = factor(variable, c("Pct_of_HLADRpos_pDC_Ki67pos",
                                              "Pct_of_HLADRpos_cDC1_Ki67pos",
                                              "Pct_of_HLADRpos_cDC2_Ki67pos",
                                              "Pct_of_HLADRpos_cDC_wo_cDC1orcDC2_Ki67pos",
                                              "Pct_of_HLADRpos_Non_classical_Mo_Ki67pos",
                                              "Pct_of_HLADRpos_Intermediate_Mo_Ki67pos",
                                              "Pct_of_HLADRpos_Classical_Mo_minusDC3_Ki67pos",
                                              "Pct_of_HLADRpos_DC3_Ki67pos"))) %>%
  # sort
  ungroup() %>%
  dplyr::arrange(compartment,Day,variable) 
 
  
  

### To name output files
contrast_name <- "FigS4B_mean_of_props"

list_of_stats[[paste0(contrast_name)]] <- curr_phen_long %>%
  dplyr::select(-comp_time)

num_columns_for_plot <- 4

list_of_plots <- list()
#collect phen_to_plots
phen_to_plots <- data.frame()

## for each comp_time
for(i in unique(curr_phen_long$comp_time)){
 phen_to_plot <- curr_phen_long %>%
    dplyr::filter(comp_time == i) %>%
    ## add in labels
    dplyr::arrange(compartment,variable) %>%
    dplyr::mutate(label = paste0(signif(mean,2),"%"),
                  ### add positions https://wilkelab.org/practicalgg/articles/bundestag_pie.html
                  cumsum_mean = cumsum(mean),
                  sum_mean = sum(mean),
                  end_angle = 2*pi*cumsum(mean)/sum(mean),   # ending angle for each pie slice
    start_angle = lag(end_angle, default = 0),   # starting angle for each pie slice
    mid_angle = 0.5*(start_angle + end_angle),   # middle of each pie slice, for the text label
    # horizontal and vertical justifications depend on whether we're to the left/right
    # or top/bottom of the pie
    hjust = ifelse(mid_angle > pi, 1, 0),
    vjust = ifelse(mid_angle < pi/2 | mid_angle > 3*pi/2, 0, 1),
                  ## add in positions for polar coord https://r-charts.com/part-whole/pie-chart-labels-outside-ggplot2/
                  # csum = rev(cumsum(rev(mean))), 
                  # pos = mean/2 + lead(csum, 1),
                  # pos = if_else(is.na(pos), mean/2, pos),
                  ### add explosion
                  focus = ifelse(mean < 10, 0.04, 0)) %>%
    ## remove rows with mean = 0
    dplyr::filter(mean != 0) 
 
 if(dim(phen_to_plots)[1] == 0){
   phen_to_plots <- phen_to_plot
 }else{
   phen_to_plots <- rbind(phen_to_plots,phen_to_plot)
 }
  
 ### set colors
  color_df <- data.frame("variable"=unique(curr_phen_long$variable),

                         "fill"=c("#D6D6D6","#0533FF","#D783FF","#73FCD6",
                                  "#FFFC79","#038F01","#FF7E79","#FF2501"),
                         "color"=c("grey50","#0533FF","#D783FF","#59A38E",
                                  "lightgoldenrod3","#038F01","#FF7E79","#FF2501"))
  

  #### base plot
  p1 <- phen_to_plot %>%
    ## plot
    ggplot() + 
    scale_fill_manual(values=c(color_df$fill[which(color_df$variable %in% unique(phen_to_plot$variable))])) + 
    scale_color_manual(values=c(color_df$color[which(color_df$variable %in% unique(phen_to_plot$variable))])) +
    ggforce::geom_arc_bar(
                 aes(x0 = 0, y0 = 0, r0 = 0, r = 0.7, amount = mean,
     fill = variable, explode = focus),
     data=phen_to_plot, stat = 'pie', size = 0.1)+
    
    geom_text(
    aes(
      x = (.75+focus) * sin(mid_angle),
      y = (.75+focus) * cos(mid_angle),
      label = label, color = variable,
      hjust = hjust, vjust = vjust
    ),
    size = 10/.pt,
    show.legend = FALSE ## no a in the legend
  ) +
    coord_fixed() +
    
    # geom_bar(stat="identity", width=1, color="black", size = 0.1) +
  #coord_polar("y", start=0, direction = -1)+
      theme_void() + 
    ### add labels
      # geom_text_repel(aes(x = 0, y = pos, 
      #                     label = label), 
      #                 size = 3, nudge_x = 1)+
    ggtitle(str_replace_all(i,"__",", "))+
  theme(plot.title = element_text(hjust = 0.5),
        ) 
  # plot(p1)

  list_of_plots[[i]] <- p1 + theme(legend.position="none")
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

### get the legend
#https://stackoverflow.com/questions/13649473/add-a-common-legend-for-combined-ggplots
g_legend<-function(a.gplot){
  tmp <- ggplot_gtable(ggplot_build(a.gplot))
  leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
  legend <- tmp$grobs[[leg]]
  return(legend)}
### add empty filler
list_of_plots[[length(list_of_plots)+1]]<-ggplot()+theme_void()
list_of_plots[[length(list_of_plots)+2]]<-g_legend(p1) ### grabs from the l
do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Figure 6

GC populations

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig6"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig6BE"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

Fig 6B

varcomp_to_plot <- c("LN__CD4_Tfh_Ki67pos","LN__GC_CD4_Tfh_Ki67pos",
                     "LN__gCD8_Tfh_Ki67pos","LN__B_Cells_Bcl6pos_Pct_B_cells")

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 6C

spearman correlation vs self

########## generate "dataset" format - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_to_use %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log2_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


analyte_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_to_use %>%
  ### remove data columns
  dplyr::select(-all_of(analyte_columns))

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D3-D1","D8-D1","D22-D1")

contrast_name <- "Fig6C_GC"

### for CorrVsSelf
set_names_to_use <- c(paste0(c("D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC_withinCP"))
vars_for_selfCorr <- varcomp_to_plot

corr_tag_name <- "CorrVsSelf"
set_name_labels <- data.frame("set_name"=set_names_to_use,
                              "label"=c("D8vsD1 (top)", "D22vsD1 (bottom)"))

fullcombo_df_long <- data.frame()
for(set_name in set_names_to_use){
  
  ######## calculate correlation
  cor_method <- "spearman"
  
  
curr_ds <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  ### trim to only the vars desired
  dplyr::filter(Gene %in% vars_for_selfCorr) %>%
  column_to_rownames("Gene")

  curr_df <- data.frame(t(curr_ds)) %>%
    
    ### trim to only tp of interest
    rownames_to_column("sampleName") %>%
    dplyr::filter(grepl(str_before_first(set_name,"_"),sampleName)) %>%
    ### be sure sorted
    dplyr::arrange(sampleName) %>%
    column_to_rownames("sampleName") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)
  

    
    ## generate cor matrix 
  cor_mat <- cor(as.matrix(signif(curr_df,7)), method = cor_method, use="pa")
 
  ## raw p
  cor_test_mat <- data.frame(matrix(NA, nrow=dim(curr_df)[2], ncol=dim(curr_df)[2]))
  rownames(cor_test_mat) <- colnames(curr_df)
  colnames(cor_test_mat) <- colnames(curr_df)
  
  for(curr1 in colnames(curr_df)) {
    for(curr2 in colnames(curr_df)) {
      cor_test_mat[curr1, curr2] <- 
        cor.test(as.matrix(signif(curr_df,7))[,curr1], 
                 as.matrix(signif(curr_df,7))[,curr2], 
                 method = cor_method, use="pa")$p.value
    }
  }
   
 ## don't bother adjusting since won't be symmetric 
  # cor_test_adj_mat <- cor_test_mat #duplicate dataframe to replace values
  # for(curr_assay in colnames(curr_df1)) {
  #   cor_test_adj_mat[curr_assay,] <- p.adjust(cor_test_mat[curr_assay,], method = "BH")
  # }

 
    
    
    raw_or_adj <- "raw"
    ann_pval_level <- 0.05
    trim_pval_level <- 0.05
     
          cat(" ", paste0(raw_or_adj," p < ", trim_pval_level),"\n")
    cat("\n")
    
    
    cat(" ", paste0("untrimmed = all analytes"),"\n")
    cat("\n")
    
    cat(" ", paste0("trimmed = only analytes with at least one association p < ",trim_pval_level),"\n")
    cat("\n")
    cat(" ", paste0("p-value for stars annotation < ",ann_pval_level),"\n")
    cat("\n")
    
    
    cat("\n")
    cat(' \n \n') ### this is the key!
    
    order <- "alphabet"
  
  vars_ann <- data.frame(cor_mat) %>%
    rownames_to_column() %>%
    arrange(rowname)
  ordered_vars_to_use <- vars_ann$rowname
  
  
  ### need this in case have only since variable being compared        
  cor_mat_for_plot <- as.matrix(cor_mat[ordered_vars_to_use,ordered_vars_to_use])
  colnames(cor_mat_for_plot) <- ordered_vars_to_use
  rownames(cor_mat_for_plot) <- ordered_vars_to_use
  

    cor_test_mat_for_plot <- as.matrix(cor_test_mat[ordered_vars_to_use,ordered_vars_to_use])
   rownames(cor_test_mat_for_plot) <- ordered_vars_to_use
  colnames(cor_test_mat_for_plot) <- ordered_vars_to_use 
  #### set diag to 1, so won't be sig
  diag(cor_test_mat_for_plot) <- 1
  
  legend_ratio = 0.3
  text_size = 0.3
  

     cat("#### ", paste0(set_name_labels$label[which(set_name_labels$set_name == set_name)]),"\n")
    cat("\n")
    
  
    
  
  if(length(which(cor_test_mat_for_plot <  as.numeric(ann_pval_level))) == 0){
    
    
    corrplot::corrplot(cor_mat_for_plot, method = 'square', 
                       is.corr = TRUE, #since symmetric
                        type = 'lower', diag = FALSE,
                       p.mat = cor_test_mat_for_plot,
                       sig.level = as.numeric(ann_pval_level), ## doesn't work if none are sig
                       #sig.level = 0.1,#set p < 0.05 to blank
                       insig='pch',
                       pch.cex = 1.2, pch.col = 'black',tl.col = 'black',tl.cex=text_size,
                       ##legend
                       cl.ratio = legend_ratio, #cl.cex = 2,
                       
                       #addCoef.col ='grey80', 
                       #number.cex = 0.5, #add corr coeff to sig circles
                       order = order, ## CANNOT ORDER ASSYMMETRIC
                       col = rev(COL2('PuOr',1000)), ##make orange positive
                       col.lim = c(-1,1) ##set limits to be the same across plots
                       # title = vaccine_groups[i], mar=c(0,0,1,0) # http://stackoverflow.com/a/14754408/54964
    )
    
    
    cat("\n")
    cat(' \n \n') ### this is the key!
    
   
    
  }else{
    #### otherwise can use the stars
    corrplot::corrplot(cor_mat_for_plot, method = 'square', 
                       is.corr = TRUE, #since symmetric
                        type = 'lower', diag = FALSE,
                    p.mat = cor_test_mat_for_plot,
                       sig.level = as.numeric(ann_pval_level), 
                       insig='label_sig',  #set p < 0.05 to blank
                       pch.cex = 1.2, pch.col = 'white',tl.col = 'black',tl.cex=text_size,
                       ##legend
                       cl.ratio = legend_ratio, #cl.cex = 2,
                       #addCoef.col ='grey80', 
                       #number.cex = 0.5, #add corr coeff to sig circles
                       order = order, ## CANNOT ORDER ASSYMMETRIC
                       col = rev(COL2('PuOr',1000)), ##make orange positive
                       col.lim = c(-1,1) ##set limits to be the same across plots
                       # title = vaccine_groups[i], mar=c(0,0,1,0) # http://stackoverflow.com/a/14754408/54964
    )
    
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
  }
   
    cat("\n")
    cat(' \n \n') ### this is the key!
    
    
    
    ######### save the stats
       ### write function for converting the cor_mat to long format df
    corr_to_longSELF <- function(mat){
      data.frame(mat) %>%
        rownames_to_column("celltype1") %>%
        tidyr::pivot_longer(!celltype1, names_to = "celltype2", values_to = "value") %>%
        dplyr::mutate(celltype1_vs_2 = paste0(celltype1,"_vs_",celltype2))
      
    }

    
    
    
    
    #### convert to long
    cor_long <-  corr_to_longSELF(cor_mat)   
    if(raw_or_adj == "raw"){
      p_long <-  corr_to_longSELF(cor_test_mat)     
    }else{
      p_long <-  corr_to_longSELF(cor_test_adj_mat)     
    }
    
    #### combine cor and p long dfs
    curr_df_long <- dplyr::left_join(cor_long,p_long, by = c("celltype1","celltype2","celltype1_vs_2")) %>%
      ### rename values to be useful
      dplyr::rename(corr_coef=value.x,p=value.y) %>%
      ### add column with set_name
      dplyr::mutate(timepoint = paste0(str_replace_all(set_name,"-","vs")))

  
  ### add to fullcombo
  fullcombo_df_long <- rbind(fullcombo_df_long,curr_df_long)

    
   
}

raw p < 0.05

untrimmed = all analytes

trimmed = only analytes with at least one association p < 0.05

p-value for stars annotation < 0.05

D8vsD1 (top)

raw p < 0.05

untrimmed = all analytes

trimmed = only analytes with at least one association p < 0.05

p-value for stars annotation < 0.05

D22vsD1 (bottom)

cat("\n")

cat(' \n \n') ### this is the key!

  CorrName <- paste0(contrast_name,"__",corr_tag_name)
### add to stats master list
list_of_stats[[paste0(CorrName)]] <- fullcombo_df_long %>% 
  dplyr::relocate(celltype1, celltype2, corr_coef, p, timepoint)

Fig 6D

##### set prior for log2 (0 for MSD, 1e-4 for cell pops, 1 for IHC)
prior_to_useIHC <- 1

##### select data
curr_df_long <- list_of_datasets[["IHC"]] %>%
  ### filter to only vars of interest = GC zone
  dplyr::filter(zone == "GC") %>%
  ### add log10
  dplyr::mutate(celltype_zone = paste0(celltype,"_",zone),
                value_plus = value_raw + as.numeric(prior_to_useIHC),
                value_log10 = log10(value_plus),
                raw_value = NULL,
                ### factorize time
                Day = factor(Day, c("D1","D3","D8","D22"))

                ) 



## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig6D_IHC"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 2

IHC differences between timepoints (within a zone)

\(log10(value + 1)_{zoneXcelltypeY}\) ~ Time + (1|MacaqueID)

Contrasts within D3, D8 D22 are only for LION treatments, since CTRLs have just one timepoint

### if want to trim which p values on plot, these are are all
contrasts_to_keep <- c("vsD1","vsD3","vsD8")
treatments_to_keep <- c("LION","CTRL")
### Scatterplots by time/zone

#### read in full pvals
pval_df_full <- list_of_stats[[paste0(contrast_name,"_lmerTestByTime")]]

### trim to contrasts
  pval_df_trimmed1 <- pval_df_full %>%
    dplyr::filter(grepl(paste(contrasts_to_keep, collapse = '|'),contrast)) %>%
    dplyr::filter(!is.na(lmer_est))

#### for each celltype "TAB", plot the median value (across replicates) for each animal/variable/zone/timepoint

cz_tabs <- unique(pval_df_trimmed1$variable)

 y_scale_df <- data.frame("celltype"=c("CD4","CD8"),
                          "color"=c("#08F8F9","#F942F1"),
                             "max"=c(5,5),
                          "min"=c(0,0),
                             "to_plot"=c(11,11))
 list_of_plots <- list()
 ### for each celltype/zone
for(curr_tab in cz_tabs){
  ## first trim pvals to the celltype "TAB"
  pval_df_trimmed <- pval_df_trimmed1[grepl(curr_tab,pval_df_trimmed1$variable),] 
  curr_celltype <- str_before_first(curr_tab,"_")
  curr_zone <- str_after_first(curr_tab,"_")
  
  ##### also trim the timepoint_comparison_df
curr_timepoint_comparison_df <- timepoint_comparison_df %>%
  dplyr::filter(grepl(paste(contrasts_to_keep, collapse = '|'),name))

#### put pvals in format for plotting, tribbles
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$variable)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(curr_timepoint_comparison_df)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               curr_timepoint_comparison_df[g,"time1"],curr_timepoint_comparison_df[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),"lmer_praw"],
                               pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),"lmer_est"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(curr_tab)]] <- pval_df
}




#### prep phen_to_plot
#### add additional factorized vars
phen_to_plot <- curr_df_long %>%
  ### filter to only lion
  dplyr::filter(treatment %in% treatments_to_keep) %>%
  ### filter variables
  dplyr::filter(celltype_zone %in% pval_df_trimmed$variable) %>%
  ### factorize
  dplyr::mutate(zone = factor(zone, c("GC","noGC","TCZ")),
                celltype = factor(celltype, c("CD4","CD8"))) %>%
     dplyr::arrange(MacaqueID)

plot_sig_only <- "no"

  
   size_for_points <- 1
   stroke_for_points <- 0.4
   dodge_for_points <- 0.9


  max_y <- y_scale_df$max[which(y_scale_df$celltype == curr_celltype)] #### set this to be consistent for each celltype

     min_y <- y_scale_df$min[which(y_scale_df$celltype == curr_celltype)]
   maxy_y_to_plot <-  y_scale_df$to_plot[which(y_scale_df$celltype == curr_celltype)]

  
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(curr_timepoint_comparison_df),
                          length(which(pval_df_trimmed[i,] %>% dplyr::select(-variable) < as.double(params$pCutoff))) #### fix
  )
  
  color_df <- data.frame("zone"=c("GC","noGC","TCZ"),
                       "shape" = c(21, 22, 24))



  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = value_log10, 
                                shape = zone, 
                                fill = celltype#,
                               # group = MacaqueID
                                )) + 
    scale_fill_manual(values=y_scale_df$color[which(y_scale_df$celltype == curr_celltype)]) +  
    scale_color_manual(values=y_scale_df$color[which(y_scale_df$celltype == curr_celltype)]) +  
    scale_shape_manual(values=color_df$shape[which(color_df$zone == curr_zone)]) +
    #scale_shape_manual(values=shapes_df$shape_for_plot[which(shapes_df$animal %in% unique(phen_to_plot$MacaqueID))]) +
#geom_line(color="grey",linetype = "dashed")+
    
    geom_beeswarm(aes(
      #fill=zone,
      #color=zone,
      shape=zone,
    ), fill="white", color="black",size = size_for_points, stroke = stroke_for_points, alpha = 0.9,corral = "random", corral.width = 0.5,
    dodge.width = dodge_for_points)+ ## makes side by side groups
       
    # geom_point(aes(
    #   fill=zone,
    #   color=zone,
    #   shape=zone,
    # ), size = 3, stroke = 0.8, alpha = 0.9)+ ## makes side by side groups
    #geom_text_repel() + 
    scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot), breaks = seq(min_y,maxy_y_to_plot,1))+
    
    #rremove("x.axis")+  
    theme_bw()+
    labs(y = paste0(curr_celltype," ",curr_zone," for all animals (log10)")#,
         #title = paste0(i)
    )+
    theme(legend.position="right", legend.box = "vertical",
          #axis.ticks.x = element_blank(),
          #axis.text.x = element_blank(),
          plot.title = element_text(hjust = 0.5), #centers title
          panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
          panel.border = element_blank(), axis.line = element_line(colour = "black")
    )

  ##### add p values

 
  ### for each zone
 
     first_interval_to_use <- 0.05
      interval_to_use <- 0.15
    

           pvals_for_plot <- list_of_pvals[[curr_tab]] %>% 
                           dplyr::mutate(zone=paste0(curr_zone),
                                         celltype=paste0(curr_celltype),
                                         MacaqueID="p") %>% 
                        # add_x_position(x="Day") %>% ### does it in alphabetical order
                         rownames_to_column() %>% 
                         dplyr::mutate(xmin = ifelse(group1 == "D1",1,
                                                     ifelse(group1 == "D3",2,3)), 
                                       xmax = ifelse(group2 == "D3",2,
                                                     ifelse(group2 == "D8",3,4
                                                            )), 
                                       y.position = c(max_y+first_interval_to_use*range_y+interval_to_use*as.numeric(rowname)*range_y))
        
    
        for(curr_row in 1:nrow(pvals_for_plot)){
            p <- p +
       
         
    stat_pvalue_manual(pvals_for_plot[curr_row,],
                       label = "p",size = 3.5,tip.length = 0.01,
                       bracket.size = 0.2,
                       
                       color = ifelse(pvals_for_plot$dir[curr_row] == "up","red",
                                                      ifelse(pvals_for_plot$dir[curr_row] == "down","blue",
                                                             ifelse(pvals_for_plot$dir[curr_row]== "not_sig","black","grey"))))
    
        }
      
      
      #### for lumped together add the median and quantiles
    p <- p+
      geom_violin(data = phen_to_plot, aes(x = Day, y = value_log10, fill = celltype), draw_quantiles = c(0.25, 0.5, 0.75), color = "grey40", alpha = 0.2,
                  inherit.aes = FALSE, scale = "width", adjust =0.5)+
           # scale_color_manual(values = c("purple4","orange","grey60","black","black"))+
      stat_summary(fun="mean", geom="crossbar", aes(color = "mean"), color = "black", linetype = "11", size = 0.3) +
      stat_summary(fun="median", geom="crossbar",aes(color = "median"), color = "black", linetype = "solid", size = 0.3)
 
     
      # plot(p)
  
cat('\n')
cat(' \n \n') ### this is the key! 
  

     
     
     

   

  
cat('\n')
cat(' \n \n') ### this is the key! 
  
  
 # plot(p)
  list_of_plots[[paste0(curr_tab)]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 6E

varcomp_to_plot <- c("LN__CD4_Tfh_CXCR3neg_Ki67pos","LN__CD4_Tfh_CXCR3pos_Ki67pos",
                     "LN__GC_CD4_Tfh_CXCR3neg_Ki67pos","LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                     "LN__CD8_Tfh_CXCR3neg_Ki67pos","LN__CD8_Tfh_CXCR3pos_Ki67pos")

#### reset contrast name to full p values
contrast_name <- "Fig6BE"

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 6F

Spearman correlation between MSD D2vsD1 and GC populations

########## generate "dataset" format - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_to_use %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log2_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


analyte_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_to_use %>%
  ### remove data columns
  dplyr::select(-all_of(analyte_columns))

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D3-D1","D8-D1","D22-D1")

contrast_name <- "Fig6FG_GC"

raw_or_adj <- "raw"

### which dataset levels/FC, trimmed to the tab
curr_ds <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  column_to_rownames("Gene")

###### MSD = D2-D1 only log2FC
curr_df2 <- data.frame(t(list_of_diff_counts[[paste0("Fig3C_MSD_V4_D2vsD1")]])) %>%
  ### make row 1 colnames
  janitor::row_to_names(row_number = 1) %>%
  ### trim to only D2-D1
  rownames_to_column("sampleName") %>%
  dplyr::filter(grepl("D2-D1",sampleName)) %>%
  ### be sure sorted
  dplyr::arrange(sampleName) %>%
  column_to_rownames("sampleName") %>%
  #### then remove columns with all NA
  select_if(~ !all(is.na(.))) %>%
  ### be sure numeric
  mutate_if(is.character, as.numeric)
  

set_names_to_use <- c(paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC"))

corr_tag_name <- "CorrVsMSD"

 ######## calculate correlation
  cor_method <- "spearman"
  
  CorrName <- paste0(contrast_name,"__",corr_tag_name)
    
  #### generate empty dataframe for combo
fullcombo_df_long <- data.frame()


  for(set_name in set_names_to_use){
   curr_df1 <- data.frame(t(curr_ds)) %>%
    
    ### trim to only tp of interest
    rownames_to_column("sampleName") %>%
    dplyr::filter(grepl(str_before_first(set_name,"_"),sampleName)) %>%
    ### be sure sorted
    dplyr::arrange(sampleName) %>%
    column_to_rownames("sampleName") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ## remove rows with all NAs
    filter(if_any(everything(), ~ !is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)
  

    ## generate cor matrix 
  cor_mat <- cor(as.matrix(signif(curr_df1,7)), as.matrix(signif(curr_df2,7)), method = cor_method, use="pa")
 
  ## raw p
  cor_test_mat <- data.frame(matrix(NA, nrow=dim(curr_df1)[2], ncol=dim(curr_df2)[2]))
  rownames(cor_test_mat) <- colnames(curr_df1)
  colnames(cor_test_mat) <- colnames(curr_df2)
  
  for(curr1 in colnames(curr_df1)) {
    for(curr2 in colnames(curr_df2)) {
      cor_test_mat[curr1, curr2] <- 
        cor.test(as.matrix(signif(curr_df1,7)[,curr1]), 
                 as.matrix(signif(curr_df2,7)[,curr2]), 
                 method = cor_method, use="pa")$p.value
    }
  }
  
 ## adj p for multiple comparisons by BH over cytokine assays
  cor_test_adj_mat <- cor_test_mat #duplicate dataframe to replace values
  for(curr_assay in colnames(curr_df1)) {
    cor_test_adj_mat[curr_assay,] <- p.adjust(cor_test_mat[curr_assay,], method = "BH")
  }


  
    ### write function for converting the cor_mat to long format df
    corr_to_long <- function(mat){
      data.frame(mat) %>%
        rownames_to_column("celltype") %>%
        tidyr::pivot_longer(!celltype, names_to = "assay", values_to = "value") %>%
        dplyr::mutate(celltype_assay = paste0(celltype,"_vs_",assay))
      
    }

    
    
    
    
    #### convert to long
    cor_long <-  corr_to_long(cor_mat)   
    if(raw_or_adj == "raw"){
      p_long <-  corr_to_long(cor_test_mat)     
    }else{
      p_long <-  corr_to_long(cor_test_adj_mat)     
    }
    
    #### combine cor and p long dfs
    curr_df_long <- dplyr::left_join(cor_long,p_long, by = c("celltype","assay","celltype_assay")) %>%
      ### rename values to be useful
      dplyr::rename(corr_coef=value.x,p=value.y) %>%
      ### add column with set_name
      dplyr::mutate(timepoint = paste0(str_replace_all(set_name,"-","vs")))

  
  ### add to fullcombo
  fullcombo_df_long <- rbind(fullcombo_df_long,curr_df_long)
  
  }

### add to stats master list
list_of_stats[[paste0(CorrName)]] <- fullcombo_df_long %>% 
  dplyr::relocate(celltype, assay, corr_coef, p, timepoint)

barbara_order <- c("LN__CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__CD8_Tfh_CXCR3pos_Ki67pos",
                   "LN__CD8_Tfh_CXCR3neg_Ki67pos",
                   "LN__Bcells_GC_BCL6pos_ki67pos_PctBCells")
 

## only for the Intervals
timepoints_to_use <- paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC")
TP_name <- "Intervals"

corr_tag_name <- "CorrVsMSD" 
raw_or_adj <- "raw"
cor_method <- "spearman"
trim_pval_level <- 0.05
all_or_selected <- "selected"
assay_arrangement <- c("CCL2.MCP.1","CCL3.MIP.1a","CCL4.MIP.1b","CCL8.MCP.2",
                       "CXCL10.IP.10","CXCL11.ITAC","M.CSF",
                       "CCL26.Eotaxin.3","CCL13.MCP.4","CXCL12.SDF.1a",
                       "FLT3L","IL.7","CX3CL1.Fractalkine") ### if trimming/ordering assays


### set pos/neg/both
directions_to_use <- c("pos")
assays_to_exclude <- c("")

raw p < 0.05

#### make df for plotting 
df_long1 <- fullcombo_df_long %>%
  ### generate dir column
  dplyr::mutate(dir = ifelse(p < trim_pval_level,
                             ifelse(corr_coef > 0, "pos","neg"),
                             "notSig"),
                ### break out compartment
                compartment = str_before_first(celltype,"__"),
                celltype = str_after_first(celltype,"__"),
                celltype_assay = str_after_first(celltype_assay,"__")) %>%
  ### remove the corr and p vals for pivoting
  dplyr::select(-c(corr_coef,p)) 



###### set up inital dataframe
## trim to only sig
bub_df2 <- df_long1[which(df_long1$dir %in% directions_to_use ),] %>%
  ### also remove any analytes_to_exclude
  dplyr::filter(!assay %in% assays_to_exclude) %>%
  dplyr::mutate(timepoint = factor(str_before_first(timepoint,"_"),str_replace_all(str_before_first(timepoints_to_use,"_"),"-","vs")),
                dir = factor(dir, directions_to_use),
                ##### add compartment to celltype
                comp_celltype = paste0(compartment,"__",celltype)) 


#### remove timepoint suffix from celltype (only needed for vs Abs/NAbs)
cleaned_celltypes <- bub_df2$comp_celltype
for(tu in timepoints_to_use){
  cleaned_celltypes <- str_remove_all(cleaned_celltypes,paste0("__",str_replace_all(tu,"-","vs")))
}

bub_df2$celltype <- cleaned_celltypes




  #### if selected tab combos, keep cell pops on the x-axis

      
         bub_df_bothordered <- bub_df2 %>%
           #### trim to desired assays
           dplyr::filter(assay %in% assay_arrangement) %>%
      ### remove factor levels for assay
      dplyr::mutate(assay = factor(as.character(assay),assay_arrangement),
                    ### also breakout for celltype
                    celltype_only = str_after_first(celltype,"__"),
                    compartment = factor(compartment, c("blood","LN"))) %>%   
      dplyr::arrange(assay) %>%
      #### trim to barbara_order celltypes
      dplyr::filter(celltype %in% barbara_order) %>%
      ## lock in factor levels for assay (rev for y axis since want it to be reversed)
      dplyr::mutate(assay = factor(assay, rev(unique(assay))),
                    ### now get celltype order
                    celltype = factor(celltype, intersect(barbara_order,unique(celltype)))) %>%
           #### now arrange by timepoint to be able to see bulls eyes
           dplyr::arrange(desc(timepoint))
    
 
      ###### reversed axes!!!
    bba <- 
      ggplot(bub_df_bothordered, aes(x = celltype, y = assay))+
      geom_point(aes(size = timepoint, color = timepoint))+
      #scale_alpha_manual(values = alpha_values)+
      scale_size_manual(values = c(3,6,10))+
      scale_color_manual(values = c("#262626","#7F7F7F","#BFBFBF"))+
      theme_bw()+
      ### make legend symbol bigger & fix color 
      # guides(color = guide_legend(override.aes = list(size = 5)),
      #        size = guide_legend(override.aes = list(color = leg_color)))+
      ## remove gridlines
      theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            ## rotate axis labels
            axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
  
  
    plot(bba)

  cat("\n")

    cat(' \n \n') ### this is the key!

      ### last plot in loop doesn't appear unless add this
  
    
 
 
  cat("\n")

    cat(' \n \n') ### this is the key!

    ###### need to fix why this plots in Rstudio but not renedered

Fig 6G

Spearman correlation between Abs & GC populations

Compare each cell population to Ab levels combining the following “paired” timepoints:

Cell pop D1 & Ab D1

Cell pop D3/4 & Ab D8

Cell pop D8 & Ab D15

Cell pop D22 & Ab D22

curr_figure_name <- "Antibody"
##### select data
data_Abs <- list_of_datasets[["Abs"]] %>%
  ### filter to only vars of interest (removes NAb)
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) 

########## generate "dataset" format for Abs - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_Abs %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log10_value = log10(raw_value), ### no prior
                raw_value = NULL,
                ### remove raw from analyte
                analyte = str_replace_all(analyte,"_raw","_log10")) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log10_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


analyte_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_Abs[,c("sampleName","MacaqueID","Day")]

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D8-D1","D15-D1","D22-D1")

contrast_name <- "Fig6G_Abs"

vars_for_AbCorr <-c("LN__CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__CD8_Tfh_CXCR3neg_Ki67pos",
                   "LN__CD8_Tfh_CXCR3pos_Ki67pos",
                   "LN__Bcells_GC_BCL6pos_ki67pos_PctBCells")
### load GC pops
contrast_name <- "Fig6FG_GC"
curr_ds <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  ### trim to only the vars desired
  dplyr::filter(Gene %in% vars_for_AbCorr) %>%
  column_to_rownames("Gene")


### load Abs
contrast_name <- "Fig6G_Abs"
curr_ds_abs <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  column_to_rownames("Gene") %>%
  t() %>%
  data.frame() %>%
  #### remove rows with all NA
  filter(if_any(everything(), ~ !is.na(.))) %>%
  ### fix names
  rownames_to_column("Animal_TP") %>%
  dplyr::mutate(#Animal_TP = str_replace_all(Animal_TP,"-","vs"),
                  #### also unify with cell pops
                Animal_TP = str_replace_all(Animal_TP,"D8-D1","D3-D1"),
                Animal_TP = str_replace_all(Animal_TP,"D15-D1","D8-D1")) 
  


#### set new for the stats
contrast_name <- "Fig6G_GC"
corr_tag_name <- "CorrVsAbPaired4TPs"
raw_or_adj <- "raw"

set_name <- "Intervals"
  ######## calculate correlation
  cor_method <- "spearman"
  
  #### rows = animal (trimmed to tp "set_name"), columns = vars
  curr_df1 <- data.frame(t(curr_ds)) %>%
    ## remove rows with all NAs
    filter(if_any(everything(), ~ !is.na(.))) %>%    
    ### be sure sorted
    rownames_to_column("sampleName") %>%
    dplyr::arrange(sampleName) %>%
    column_to_rownames("sampleName") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)

  
  ### rows = animal, columns = vars, including day/interval
  curr_df2 <- data.frame(curr_ds_abs) %>%
    ### be sure sorted
    dplyr::arrange(Animal_TP) %>%
    column_to_rownames("Animal_TP") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)
  
  
     CorrName <- paste0(contrast_name,"__",corr_tag_name)
    
     ## generate cor matrix 
  cor_mat <- cor(as.matrix(signif(curr_df1,7)), as.matrix(signif(curr_df2,7)), method = cor_method, use="pa")
 
  ## raw p
  cor_test_mat <- data.frame(matrix(NA, nrow=dim(curr_df1)[2], ncol=dim(curr_df2)[2]))
  rownames(cor_test_mat) <- colnames(curr_df1)
  colnames(cor_test_mat) <- colnames(curr_df2)
  
  for(curr1 in colnames(curr_df1)) {
    for(curr2 in colnames(curr_df2)) {
      cor_test_mat[curr1, curr2] <- 
        cor.test(as.matrix(signif(curr_df1,7)[,curr1]), 
                 as.matrix(signif(curr_df2,7)[,curr2]), 
                 method = cor_method, use="pa")$p.value
    }
  }
  
 ## adj p for multiple comparisons by BH over cytokine assays
  cor_test_adj_mat <- cor_test_mat #duplicate dataframe to replace values
  for(curr_assay in colnames(curr_df1)) {
    cor_test_adj_mat[curr_assay,] <- p.adjust(cor_test_mat[curr_assay,], method = "BH")
  }


  
    ### write function for converting the cor_mat to long format df
    corr_to_long <- function(mat){
      data.frame(mat) %>%
        rownames_to_column("celltype") %>%
        tidyr::pivot_longer(!celltype, names_to = "assay", values_to = "value") %>%
        dplyr::mutate(celltype_assay = paste0(celltype,"_vs_",assay))
      
    }

    
    
    
    
    #### convert to long
    cor_long <-  corr_to_long(cor_mat)   
    if(raw_or_adj == "raw"){
      p_long <-  corr_to_long(cor_test_mat)     
    }else{
      p_long <-  corr_to_long(cor_test_adj_mat)     
    }
    
    #### combine cor and p long dfs
    curr_df_long <- dplyr::left_join(cor_long,p_long, by = c("celltype","assay","celltype_assay")) %>%
      ### rename values to be useful
      dplyr::rename(corr_coef=value.x,p=value.y) %>%
      ### add column with set_name
      dplyr::mutate(timepoint = paste0(str_replace_all(set_name,"-","vs")))

  
 
### add to stats master list
list_of_stats[[paste0(CorrName)]] <- curr_df_long %>% 
  dplyr::relocate(celltype, assay, corr_coef, p, timepoint)  
    
    
    raw_or_adj <- "raw"
    ann_pval_level <- 0.05
    trim_pval_level <- 0.05
   
    
    cat("\n")

    cat(' \n \n') ### this is the key!

    ######### plotting

  
  vars_ann <- data.frame(cor_mat) %>%
    rownames_to_column() %>%
    arrange(rowname)
  ordered_vars_to_use <- vars_ann$rowname
  
  ABs_ann <- data.frame(t(cor_mat)) %>%
    rownames_to_column() %>%
    arrange(rowname)
  ordered_ABs_to_use <- ABs_ann$rowname
  
  
  ### need this in case have only since variable being compared        
  cor_mat_for_plot <- as.matrix(cor_mat[ordered_vars_to_use,ordered_ABs_to_use])
  colnames(cor_mat_for_plot) <- ordered_ABs_to_use
  
  if(raw_or_adj == "raw"){
    cor_test_mat_for_plot <- as.matrix(cor_test_mat[ordered_vars_to_use,ordered_ABs_to_use])
  }else{
    cor_test_mat_for_plot <- as.matrix(cor_test_adj_mat[ordered_vars_to_use,ordered_ABs_to_use])
  }
  rownames(cor_test_mat_for_plot) <- ordered_vars_to_use
  colnames(cor_test_mat_for_plot) <- ordered_ABs_to_use          
  
  legend_ratio = 0.9
  text_size = 0.8
 
  star_size <- 3
  
  
  
  
    #### plot with the star annotation
    corrplot::corrplot(cor_mat_for_plot, method = 'square', 
                       is.corr = FALSE, #since not symmetric
                       p.mat = cor_test_mat_for_plot,
                       sig.level = as.numeric(ann_pval_level), 
                       insig='label_sig',  #set p < 0.05 to blank
                       pch.cex = star_size, pch.col = 'white',tl.col = 'black',tl.cex=text_size,
                       ##legend
                       cl.ratio = legend_ratio, #cl.cex = 2,
                       #addCoef.col ='grey80', 
                       #number.cex = 0.5, #add corr coeff to sig circles
                       #order = order, ## CANNOT ORDER ASSYMMETRIC
                       col = rev(COL2('PuOr',1000)), ##make orange positive
                       col.lim = c(-1,1) ##set limits to be the same across plots
                       # title = vaccine_groups[i], mar=c(0,0,1,0) # http://stackoverflow.com/a/14754408/54964
    )

  cat("\n")

  cat(' \n \n') ### this is the key!

Supp Figure S5

Fig S5A

contrast_name <- "Fig6BE"
data_to_useFig6 <- data_to_use
### trim to only vars to plot
variables_ordered <- c("LN__CD4_Tfh_Ki67pos","LN__GC_CD4_Tfh_Ki67pos",
                       "LN__gCD8_Tfh_Ki67pos","LN__B_Cells_Bcl6pos_Pct_B_cells")
data_to_use <- data_to_useFig6 %>%
  dplyr::select(c(sampleName,MacaqueID,Day,
                       !!!rlang::syms(variables_ordered))
  )

num_columns_for_plot <- 4

list_of_stats[["FigS5AC"]] <- data.frame("FigS5A"="stats are in the Fig6BE tab")

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),variables_ordered),
                Day = factor(str_after_first(sampleName,"__"),timepoints_x_order),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig S5B

##### set prior for log2 (0 for MSD, 1e-4 for cell pops, 1 for IHC)
prior_to_useIHC <- 1

##### select data
curr_df_long <- list_of_datasets[["IHC"]] %>%
  ### filter to only vars of interest = GC zone
  dplyr::filter(zone != "GC") %>%
  ### add log10
  dplyr::mutate(celltype_zone = paste0(celltype,"_",zone),
                value_plus = value_raw + as.numeric(prior_to_useIHC),
                value_log10 = log10(value_plus),
                raw_value = NULL,
                ### factorize time
                Day = factor(Day, c("D1","D3","D8","D22"))

                ) 



## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "FigS5B_IHC"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 2

IHC differences between timepoints (within a zone)

\(log10(value + 1)_{zoneXcelltypeY}\) ~ Time + (1|MacaqueID)

Contrasts within D3, D8 D22 are only for LION treatments, since CTRLs have just one timepoint

### if want to trim which p values on plot, these are are all
contrasts_to_keep <- c("vsD1","vsD3","vsD8")
treatments_to_keep <- c("LION","CTRL")
### Scatterplots by time/zone

#### read in full pvals
pval_df_full <- list_of_stats[[paste0(contrast_name,"_lmerTestByTime")]]

### trim to contrasts
  pval_df_trimmed1 <- pval_df_full %>%
    dplyr::filter(grepl(paste(contrasts_to_keep, collapse = '|'),contrast)) %>%
    dplyr::filter(!is.na(lmer_est))

#### for each celltype "TAB", plot the median value (across replicates) for each animal/variable/zone/timepoint

cz_tabs <- unique(pval_df_trimmed1$variable)

 y_scale_df <- data.frame("celltype"=c("CD4","CD8"),
                          "color"=c("#08F8F9","#F942F1"),
                             "max"=c(5,5),
                          "min"=c(0,0),
                             "to_plot"=c(11,11))
 list_of_plots <- list()
 ### for each celltype/zone
for(curr_tab in cz_tabs){
  ## first trim pvals to the celltype "TAB"
  pval_df_trimmed <- pval_df_trimmed1[grepl(curr_tab,pval_df_trimmed1$variable),] 
  curr_celltype <- str_before_first(curr_tab,"_")
  curr_zone <- str_after_first(curr_tab,"_")
  
  ##### also trim the timepoint_comparison_df
curr_timepoint_comparison_df <- timepoint_comparison_df %>%
  dplyr::filter(grepl(paste(contrasts_to_keep, collapse = '|'),name))

#### put pvals in format for plotting, tribbles
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$variable)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(curr_timepoint_comparison_df)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               curr_timepoint_comparison_df[g,"time1"],curr_timepoint_comparison_df[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),"lmer_praw"],
                               pval_df_trimmed[which(pval_df_trimmed$variable == i & pval_df_trimmed$contrast == curr_timepoint_comparison_df[g,"name"]),"lmer_est"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(curr_tab)]] <- pval_df
}




#### prep phen_to_plot
#### add additional factorized vars
phen_to_plot <- curr_df_long %>%
  ### filter to only lion
  dplyr::filter(treatment %in% treatments_to_keep) %>%
  ### filter variables
  dplyr::filter(celltype_zone %in% pval_df_trimmed$variable) %>%
  ### factorize
  dplyr::mutate(zone = factor(zone, c("GC","noGC","TCZ")),
                celltype = factor(celltype, c("CD4","CD8"))) %>%
     dplyr::arrange(MacaqueID)

plot_sig_only <- "no"

  
   size_for_points <- 1
   stroke_for_points <- 0.4
   dodge_for_points <- 0.9


  max_y <- y_scale_df$max[which(y_scale_df$celltype == curr_celltype)] #### set this to be consistent for each celltype

     min_y <- y_scale_df$min[which(y_scale_df$celltype == curr_celltype)]
   maxy_y_to_plot <-  y_scale_df$to_plot[which(y_scale_df$celltype == curr_celltype)]

  
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(curr_timepoint_comparison_df),
                          length(which(pval_df_trimmed[i,] %>% dplyr::select(-variable) < as.double(params$pCutoff))) #### fix
  )
  
  color_df <- data.frame("zone"=c("GC","noGC","TCZ"),
                       "shape" = c(21, 22, 24))



  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = value_log10, 
                                shape = zone, 
                                fill = celltype#,
                               # group = MacaqueID
                                )) + 
    scale_fill_manual(values=y_scale_df$color[which(y_scale_df$celltype == curr_celltype)]) +  
    scale_color_manual(values=y_scale_df$color[which(y_scale_df$celltype == curr_celltype)]) +  
    scale_shape_manual(values=color_df$shape[which(color_df$zone == curr_zone)]) +
    #scale_shape_manual(values=shapes_df$shape_for_plot[which(shapes_df$animal %in% unique(phen_to_plot$MacaqueID))]) +
#geom_line(color="grey",linetype = "dashed")+
    
    geom_beeswarm(aes(
      #fill=zone,
      #color=zone,
      shape=zone,
    ), fill="white", color="black",size = size_for_points, stroke = stroke_for_points, alpha = 0.9,corral = "random", corral.width = 0.5,
    dodge.width = dodge_for_points)+ ## makes side by side groups
       
    # geom_point(aes(
    #   fill=zone,
    #   color=zone,
    #   shape=zone,
    # ), size = 3, stroke = 0.8, alpha = 0.9)+ ## makes side by side groups
    #geom_text_repel() + 
    scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot), breaks = seq(min_y,maxy_y_to_plot,1))+
    
    #rremove("x.axis")+  
    theme_bw()+
    labs(y = paste0(curr_celltype," ",curr_zone," for all animals (log10)")#,
         #title = paste0(i)
    )+
    theme(legend.position="right", legend.box = "vertical",
          #axis.ticks.x = element_blank(),
          #axis.text.x = element_blank(),
          plot.title = element_text(hjust = 0.5), #centers title
          panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
          panel.border = element_blank(), axis.line = element_line(colour = "black")
    )

  ##### add p values

 
  ### for each zone
 
     first_interval_to_use <- 0.05
      interval_to_use <- 0.15
    

           pvals_for_plot <- list_of_pvals[[curr_tab]] %>% 
                           dplyr::mutate(zone=paste0(curr_zone),
                                         celltype=paste0(curr_celltype),
                                         MacaqueID="p") %>% 
                        # add_x_position(x="Day") %>% ### does it in alphabetical order
                         rownames_to_column() %>% 
                         dplyr::mutate(xmin = ifelse(group1 == "D1",1,
                                                     ifelse(group1 == "D3",2,3)), 
                                       xmax = ifelse(group2 == "D3",2,
                                                     ifelse(group2 == "D8",3,4
                                                            )), 
                                       y.position = c(max_y+first_interval_to_use*range_y+interval_to_use*as.numeric(rowname)*range_y))
        
    
        for(curr_row in 1:nrow(pvals_for_plot)){
            p <- p +
       
         
    stat_pvalue_manual(pvals_for_plot[curr_row,],
                       label = "p",size = 3.5,tip.length = 0.01,
                       bracket.size = 0.2,
                       
                       color = ifelse(pvals_for_plot$dir[curr_row] == "up","red",
                                                      ifelse(pvals_for_plot$dir[curr_row] == "down","blue",
                                                             ifelse(pvals_for_plot$dir[curr_row]== "not_sig","black","grey"))))
    
        }
      
      
      #### for lumped together add the median and quantiles
    p <- p+
      geom_violin(data = phen_to_plot, aes(x = Day, y = value_log10, fill = celltype), draw_quantiles = c(0.25, 0.5, 0.75), color = "grey40", alpha = 0.2,
                  inherit.aes = FALSE, scale = "width", adjust =0.5)+
           # scale_color_manual(values = c("purple4","orange","grey60","black","black"))+
      stat_summary(fun="mean", geom="crossbar", aes(color = "mean"), color = "black", linetype = "11", size = 0.3) +
      stat_summary(fun="median", geom="crossbar",aes(color = "median"), color = "black", linetype = "solid", size = 0.3)
 
     
      # plot(p)
  
cat('\n')
cat(' \n \n') ### this is the key! 
  

     
     
     

   

  
cat('\n')
cat(' \n \n') ### this is the key! 
  
  
 # plot(p)
  list_of_plots[[paste0(curr_tab)]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig S5C

contrast_name <- "Fig6BE"

### trim to only vars to plot
variables_ordered <- c("LN__CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                   "LN__GC_CD4_Tfh_CXCR3neg_Ki67pos",
                   "LN__CD8_Tfh_CXCR3pos_Ki67pos",
                   "LN__CD8_Tfh_CXCR3neg_Ki67pos")
data_to_use <- data_to_useFig6 %>%
  dplyr::select(c(sampleName,MacaqueID,Day,
                       !!!rlang::syms(variables_ordered))
  )

num_columns_for_plot <- 3

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),variables_ordered),
                Day = factor(str_after_first(sampleName,"__"),timepoints_x_order),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = Day, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_beeswarm(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 10), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Supp Figure S6

FigS6

Selected scatterplots

########## Reload data from Fig 6F
### which dataset levels/FC, trimmed to the tab
curr_ds <- list_of_diff_counts[[paste0("Fig6FG_GC")]] %>%
  column_to_rownames("Gene")

###### MSD = D2-D1 only log2FC
curr_df2 <- data.frame(t(list_of_diff_counts[[paste0("Fig3C_MSD_V4_D2vsD1")]])) %>%
  ### make row 1 colnames
  janitor::row_to_names(row_number = 1) %>%
  ### trim to only D2-D1
  rownames_to_column("sampleName") %>%
  dplyr::filter(grepl("D2-D1",sampleName)) %>%
  ### be sure sorted
  dplyr::arrange(sampleName) %>%
  column_to_rownames("sampleName") %>%
  #### then remove columns with all NA
  select_if(~ !all(is.na(.))) %>%
  ### be sure numeric
  mutate_if(is.character, as.numeric)
  

selected_pairs_df <- data.frame("var_x"=c("LN__CD4_Tfh_CXCR3pos_Ki67pos",
                                          "LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                                          "LN__CD4_Tfh_CXCR3pos_Ki67pos",
                                          "LN__GC_CD4_Tfh_CXCR3pos_Ki67pos",
                                          "LN__CD8_Tfh_CXCR3pos_Ki67pos",
                                          "LN__Bcells_GC_BCL6pos_ki67pos_PctBCells",
                                          "LN__CD8_Tfh_CXCR3pos_Ki67pos",
                                          "LN__Bcells_GC_BCL6pos_ki67pos_PctBCells"),
                                "Day_x"=c("D8-D1",
                                             "D3-D1",
                                             "D3-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D8-D1",
                                             "D8-D1"
                                             ),
                                "var_y"=c("IL.7",
                                          "FLT3L",
                                          "CXCL10.IP.10",
                                          "CXCL11.ITAC",
                                          "CXCL10.IP.10",
                                          "FLT3L",
                                          "CXCL11.ITAC",
                                          "CXCL12.SDF.1a"),
                                "Day_y"=c(rep("D2-D1",8))
)

num_columns_for_plot <- 2

### pivot curr_ds to separate timepoint from animal
df_long <- curr_ds %>%
  rownames_to_column("comp_var") %>%
  tidyr::pivot_longer(!comp_var, names_to = "sampleName", values_to = "log2FC") %>%
  ### break out MacaqueID and Day
  dplyr::mutate(Day = str_after_first(sampleName,"__"),
                Day = str_replace_all(Day,"-","vs"),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  ### add in the second df
  dplyr::full_join(data.frame(t(curr_df2)) %>%
                     rownames_to_column("comp_var") %>%
                     tidyr::pivot_longer(!comp_var, 
                                         names_to = "sampleName", values_to = "log2FC") %>%
                     ### break out MacaqueID and Day
                     dplyr::mutate(Day = str_after_first(sampleName,"__"),
                                   Day = str_replace_all(Day,"\\.","vs"),
                                   Day = str_replace_all(Day,"-","vs"),
                                   MacaqueID = str_before_first(sampleName,"__"),
                                   sampleName = NULL),
                   by = c("comp_var","Day","MacaqueID","log2FC")
  )



list_of_plots <- list()
## for each selected pair
for(i in 1:nrow(selected_pairs_df)){
  curr_x <- selected_pairs_df$var_x[i]
  curr_Day_x <- str_replace_all(selected_pairs_df$Day_x[i],"-","vs")
  curr_y <- selected_pairs_df$var_y[i]
  curr_Day_y <- str_replace_all(selected_pairs_df$Day_y[i],"-","vs")
  
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var %in% c(curr_x,curr_y)) %>%
   dplyr::filter(Day %in% c(curr_Day_x,curr_Day_y)) %>%
   ### combine var and time
   dplyr::mutate(comp_var_day = paste0(comp_var,"___",Day),
                 comp_var = NULL, Day = NULL) %>%
   tidyr::pivot_wider(names_from = comp_var_day, values_from = log2FC)
  

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = !!rlang::sym(paste0(curr_x,"___",curr_Day_x)), 
                                y = !!rlang::sym(paste0(curr_y,"___",curr_Day_y)))) + 
    geom_point()+
    stat_cor(method = "spearman")+
    xlab(paste0(curr_x,"\n",curr_Day_x))+
    ylab(paste0(curr_y,"\n",curr_Day_y))+
           
    theme_bw()+
    theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title = element_text(size = 10)) ### make smaller to fit
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Figure 7

CD8 Tcm, Tem, Tfc

Fig7A

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig7A"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl(curr_figure_name,colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID)) 


curr_phen_long  <- data_to_use %>%
  ### get average for each day/compartment
  # first make longer
  dplyr::select(-sampleName) %>%
  tidyr::pivot_longer(cols = contains("__"), names_to = "comp_var", values_to = "value") %>%
  dplyr::mutate(compartment = str_before_first(comp_var,"__"),
                variable = str_after_last(comp_var,"__"),
                Day = factor(Day, c("D1","D3","D8","D22"))) %>%
  
   
  ### collect proportion of 100% and means
  # collect total
  group_by(compartment, Day, MacaqueID) %>%
  dplyr::mutate(total_per_comp_time = sum(value)) %>%
  ungroup() %>%
  # calculate prop within each animal
  group_by(compartment, variable, Day, MacaqueID) %>%
  dplyr::mutate(prop = value / total_per_comp_time *100) %>%  
  ungroup() %>%
  # take mean across animals
  group_by(compartment, variable, Day) %>%
  dplyr::summarise(mean = mean(prop)) %>%
   # add comp_time
  dplyr::mutate(comp_time = paste0(compartment,"__",Day),
                variable = factor(variable, c("CM","EM","Naive","cTfc","Tcf1","Tcf2"))) %>%

  # sort
  ungroup() %>%
  dplyr::arrange(compartment,Day,variable) 
 
  
 



### To name output files
contrast_name <- "Fig7A_mean_of_props"

list_of_stats[[paste0(contrast_name)]] <- curr_phen_long %>%
  dplyr::select(-comp_time)

num_columns_for_plot <- 4

list_of_plots <- list()
## for each comp_time
for(i in unique(curr_phen_long$comp_time)){
 phen_to_plot <- curr_phen_long %>%
    dplyr::filter(comp_time == i)
  
 ### set colors
  color_df <- data.frame("variable"=unique(curr_phen_long$variable),

                         "fill"=c("#021993","#FF2500","#E9E9E9","#72FDFF","#8F8C01","#FF9400"),
                         "color"=c("white","white","black","grey40","black","white"))

  #### base plot
  p <- phen_to_plot %>%
    ## add in labels
    dplyr::arrange(variable) %>%
    dplyr::mutate(label = paste0(variable,"\n",signif(mean,2))) %>%
    ## plot
    ggplot(aes(x ="", y = mean, fill=variable)) + 
    scale_fill_manual(values=c(color_df$fill[which(color_df$variable %in% unique(phen_to_plot$variable))])) + 
    scale_color_manual(values=c(color_df$color[which(color_df$variable %in% unique(phen_to_plot$variable))])) +
    geom_bar(stat="identity", width=1, color="grey") +
  coord_polar("y", start=0, direction = -1)+
      theme_void() + 
    ### add labels
      geom_text(aes(label = label, color = variable), cex = 3,
            position = position_stack(vjust = 0.5))+
    ggtitle(str_replace_all(i,"__",", "))+
  theme(legend.position="none",plot.title = element_text(hjust = 0.5)) 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig7"

##### select data
data_to_use <- list_of_datasets[["CellPops"]] %>%
  ### filter to only vars of interes
  dplyr::select(c(sampleName,MacaqueID,Day,!!!rlang::syms(colnames(.)[grepl("Fig7_",colnames(.))]))) %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) %>%
  ### remove the control rows that are only there for Fig4E
  dplyr::filter(!grepl("control",MacaqueID))

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("D3","D8","D22",
            "D8","D22","D22"),
  "time2"=c("D1","D1","D1",
            "D3","D3","D8")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig7BC"
#contrast_name <- "Fig5D_Myeloid_vsD1"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

Fig 7B

varcomp_to_plot <- c("blood__CD8_CM_Ki67pos","LN__CD8_CM_Ki67pos",
                     "blood__CD8_EM_Ki67pos","LN__CD8_EM_Ki67pos",
                     "blood__CD8_CM_GzmBpos_Ki67pos","LN__CD8_CM_GzmBpos_Ki67pos",
                     "blood__CD8_EM_GzmBpos_Ki67pos","LN__CD8_EM_GzmBpos_Ki67pos")

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 7C

varcomp_to_plot <- c("blood__CD8_Tfh_Ki67pos","LN__CD8_Tfh_Ki67pos",
                     "blood__CD8_Tfh_GzmBpos","LN__CD8_Tfh_GzmBpos",
                     "blood__CD8_Tfh_GzmBpos_Ki67pos","LN__CD8_Tfh_GzmBpos_Ki67pos")

### read in degs results
degs_df <- list_of_stats[[paste0(contrast_name)]] %>%
  #### trim to only the vars in the figure panel
  dplyr::filter(rowname %in% varcomp_to_plot)


#### for hm only want the vs D1
degs_df_for_hm <- degs_df[,c("rowname",colnames(degs_df)[grepl("D1_log2FC",colnames(degs_df))])] %>%
  column_to_rownames() %>%
  ### remove rows with all na
  filter(if_any(everything(), ~ !is.na(.))) %>%
  `colnames<-`(str_remove_all(colnames(.),"-D1_log2FC")) %>%
  ## sort
  rownames_to_column() %>%
  arrange(rowname) %>%
  column_to_rownames()


timepoint_factor_order <- c("D3","D8","D22")


############### annotation colors (set number of dirs)
annotation_colors <-  list(Day = c(D3="grey40",D8="grey80",D22="grey90"),
                           dir1 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir2 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir3 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir4 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir5 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue"),
                           dir6 = c(up="red",up_p_only="lightpink",not_sig="grey95",down_p_only="skyblue",down="blue")
                           )

names(annotation_colors[["Day"]]) <- timepoint_factor_order
names(annotation_colors) <- c("Day",paste0(str_replace_all(TimePoints_to_compare$name,"-","vs"),"_dir"))


######### annotation for columns
annotation_col <- data.frame("contrast"= colnames(degs_df_for_hm)) %>%
  mutate(Day=factor(contrast, timepoint_factor_order)) %>%
  arrange(Day) %>%
  column_to_rownames("contrast")



############## add row annotation with sig change for FC

### generate contrasts
curr_contrasts <- c()
for(npt in 1:nrow(TimePoints_to_compare)){
  cont <- paste0(TimePoints_to_compare$time1[npt],"-",TimePoints_to_compare$time2[npt])
  curr_contrasts <- c(curr_contrasts,cont)
}

annotation_row1 <- data.frame("rowname"=degs_df$rowname)
for(curr_contrast in rev(curr_contrasts)){
### trim to curr contrast
  curr_pval_df <- degs_df[,c("rowname",colnames(degs_df)[grepl(curr_contrast,colnames(degs_df))])]
  ### rename column names to be generic
  orig_names <- colnames(curr_pval_df)
  simp_names <- ifelse(grepl("_",orig_names),str_after_last(orig_names,"_"),orig_names)
  colnames(curr_pval_df) <- simp_names
  curr_pval_df <- curr_pval_df %>%
    dplyr::mutate(dir = ifelse(pval < as.double(params$pCutoff),
                               ifelse(log2FC > as.double(params$FCcutoff),"up", ifelse(log2FC > 0, "up_p_only",
                                      ifelse(log2FC < (-1)*as.double(params$FCcutoff),"down",ifelse(log2FC < 0, "down_p_only","p_only")))),"not_sig"))
  curr_df <- curr_pval_df[,c("rowname","dir")] %>%
    dplyr::mutate(dir = factor(dir, c("up","up_p_only","not_sig","down_p_only","down")))
  colnames(curr_df) <- c("rowname",paste0(str_remove_all(str_replace_all(curr_contrast,"-","vs"),"V4_"),"_dir"))
  
  annotation_row1 <- dplyr::left_join(annotation_row1,curr_df, by = "rowname")
}

#### make rownames
annotation_row <- column_to_rownames(annotation_row1)
### sort to specified order
curr_ann_row <- annotation_row[varcomp_to_plot,]

mycols = diverging_hcl(n=100, palette="Blue-Red 3")
scale_range = 2
breakscale <- seq(-1*scale_range, scale_range, length=100)



  

p <- pheatmap(degs_df_for_hm[rownames(curr_ann_row),rownames(annotation_col)], scale="none", col = mycols,
                breaks=breakscale, clustering_method="ward.D2",
                #gaps_col = gaps_to_use, 
              border_color = NA,
                cluster_rows = FALSE, cluster_cols = FALSE,
                annotation_col = annotation_col,show_colnames = FALSE,
              annotation_row = curr_ann_row,
                annotation_colors = annotation_colors
)
p

  cat("\n")

cat(' \n \n') ### this is the key!

Fig 7D

Spearman correlation between MSD D2vsD1 and CD8 T CM/EM populations

########## generate "dataset" format - wide, no metadata with columns = samples (including vax, timepoint, animal), rows = analytes
##### has analyte names as rownames AND as column labeled gene
df <- data_to_use %>%
  ### remove metadata
  dplyr::select(-c(MacaqueID,Day)) %>%
  ### make longer to log2
  tidyr::pivot_longer(cols=!sampleName, names_to = "analyte", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL) %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "sampleName", values_from = "log2_value") %>%
  dplyr::mutate(Gene=analyte) %>%
  column_to_rownames("analyte") %>%
  relocate(Gene)


analyte_columns <- rownames(df)


## generate metadata only
targetfile_full <- data_to_use %>%
  ### remove data columns
  dplyr::select(-all_of(analyte_columns))

id_var_of_interest <- "sampleName"
var_of_interest <- "Day" ### matches groups(=timepoints) in contrasts, no MacaqueID
## designate covariates with Day related first, then id
covariates=c("Day","MacaqueID")

contrasts <- c("D3-D1","D8-D1","D22-D1")

contrast_name <- "Fig7DE_CD8T"
raw_or_adj <- "raw"

### which dataset levels/FC, trimmed to the tab
curr_ds <- list_of_diff_counts[[paste0(contrast_name)]] %>%
  column_to_rownames("Gene")

###### MSD = D2-D1 only log2FC
curr_df2 <- data.frame(t(list_of_diff_counts[[paste0("Fig3C_MSD_V4_D2vsD1")]])) %>%
  ### make row 1 colnames
  janitor::row_to_names(row_number = 1) %>%
  ### trim to only D2-D1
  rownames_to_column("sampleName") %>%
  dplyr::filter(grepl("D2-D1",sampleName)) %>%
  ### be sure sorted
  dplyr::arrange(sampleName) %>%
  column_to_rownames("sampleName") %>%
  #### then remove columns with all NA
  select_if(~ !all(is.na(.))) %>%
  ### be sure numeric
  mutate_if(is.character, as.numeric)
  

set_names_to_use <- c(paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC"))

corr_tag_name <- "CorrVsMSD"

 ######## calculate correlation
  cor_method <- "spearman"
  
  CorrName <- paste0(contrast_name,"__",corr_tag_name)
    
  #### generate empty dataframe for combo
fullcombo_df_long <- data.frame()


  for(set_name in set_names_to_use){
   curr_df1 <- data.frame(t(curr_ds)) %>%
    
    ### trim to only tp of interest
    rownames_to_column("sampleName") %>%
    dplyr::filter(grepl(str_before_first(set_name,"_"),sampleName)) %>%
    ### be sure sorted
    dplyr::arrange(sampleName) %>%
    column_to_rownames("sampleName") %>%
    #### then remove columns with all NA
    select_if(~ !all(is.na(.))) %>%
    ## remove rows with all NAs
    filter(if_any(everything(), ~ !is.na(.))) %>%
    ### be sure numeric
    mutate_if(is.character, as.numeric)
  

    ## generate cor matrix 
  cor_mat <- cor(as.matrix(signif(curr_df1,7)), as.matrix(signif(curr_df2,7)), method = cor_method, use="pa")
 
  ## raw p
  cor_test_mat <- data.frame(matrix(NA, nrow=dim(curr_df1)[2], ncol=dim(curr_df2)[2]))
  rownames(cor_test_mat) <- colnames(curr_df1)
  colnames(cor_test_mat) <- colnames(curr_df2)
  
  for(curr1 in colnames(curr_df1)) {
    for(curr2 in colnames(curr_df2)) {
      cor_test_mat[curr1, curr2] <- 
        cor.test(as.matrix(signif(curr_df1,7)[,curr1]), 
                 as.matrix(signif(curr_df2,7)[,curr2]), 
                 method = cor_method, use="pa")$p.value
    }
  }
  
 ## adj p for multiple comparisons by BH over cytokine assays
  cor_test_adj_mat <- cor_test_mat #duplicate dataframe to replace values
  for(curr_assay in colnames(curr_df1)) {
    cor_test_adj_mat[curr_assay,] <- p.adjust(cor_test_mat[curr_assay,], method = "BH")
  }


  
    ### write function for converting the cor_mat to long format df
    corr_to_long <- function(mat){
      data.frame(mat) %>%
        rownames_to_column("celltype") %>%
        tidyr::pivot_longer(!celltype, names_to = "assay", values_to = "value") %>%
        dplyr::mutate(celltype_assay = paste0(celltype,"_vs_",assay))
      
    }

    
    
    
    
    #### convert to long
    cor_long <-  corr_to_long(cor_mat)   
    if(raw_or_adj == "raw"){
      p_long <-  corr_to_long(cor_test_mat)     
    }else{
      p_long <-  corr_to_long(cor_test_adj_mat)     
    }
    
    #### combine cor and p long dfs
    curr_df_long <- dplyr::left_join(cor_long,p_long, by = c("celltype","assay","celltype_assay")) %>%
      ### rename values to be useful
      dplyr::rename(corr_coef=value.x,p=value.y) %>%
      ### add column with set_name
      dplyr::mutate(timepoint = paste0(str_replace_all(set_name,"-","vs")))

  
  ### add to fullcombo
  fullcombo_df_long <- rbind(fullcombo_df_long,curr_df_long)
  
  }

### add to stats master list
list_of_stats[[paste0(CorrName)]] <- fullcombo_df_long %>% 
  dplyr::relocate(celltype, assay, corr_coef, p, timepoint)

barbara_order <- c("blood__CD8_CM_Ki67pos",
                   "blood__CD8_CM_GzmBpos",
                   "LN__CD8_CM_Ki67pos",
                   "LN__CD8_CM_GzmBpos",
                   
                   
                   "blood__CD8_EM_Ki67pos",
                   "blood__CD8_EM_GzmBpos",
                   "LN__CD8_EM_Ki67pos",
                   "LN__CD8_EM_GzmBpos")
 

## only for the Intervals
timepoints_to_use <- paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC")
TP_name <- "Intervals"

corr_tag_name <- "CorrVsMSD" 
raw_or_adj <- "raw"
cor_method <- "spearman"
trim_pval_level <- 0.05
all_or_selected <- "selected"
assay_arrangement <- c("CCL2.MCP.1","CCL3.MIP.1a","CCL4.MIP.1b","CCL8.MCP.2",
                       "CXCL10.IP.10","CXCL11.ITAC","IL.15","M.CSF",
                       "CCL26.Eotaxin.3","CCL13.MCP.4","CXCL12.SDF.1a",
                       "FLT3L","IL.7","CX3CL1.Fractalkine","IFN.g") ### if trimming/ordering assays
### set pos/neg/both
directions_to_use <- c("pos")
assays_to_exclude <- c("CCL27.CTACK","IL.5")

raw p < 0.05

#### make df for plotting 
df_long1 <- fullcombo_df_long %>%
  ### generate dir column
  dplyr::mutate(dir = ifelse(p < trim_pval_level,
                             ifelse(corr_coef > 0, "pos","neg"),
                             "notSig"),
                ### break out compartment
                compartment = str_before_first(celltype,"__"),
                celltype = str_after_first(celltype,"__"),
                celltype_assay = str_after_first(celltype_assay,"__")) %>%
  ### remove the corr and p vals for pivoting
  dplyr::select(-c(corr_coef,p)) 



###### set up inital dataframe
## trim to only sig
bub_df2 <- df_long1[which(df_long1$dir %in% directions_to_use ),] %>%
  ### also remove any analytes_to_exclude
  dplyr::filter(!assay %in% assays_to_exclude) %>%
  dplyr::mutate(timepoint = factor(str_before_first(timepoint,"_"),str_replace_all(str_before_first(timepoints_to_use,"_"),"-","vs")),
                dir = factor(dir, directions_to_use),
                ##### add compartment to celltype
                comp_celltype = paste0(compartment,"__",celltype)) 


#### remove timepoint suffix from celltype (only needed for vs Abs/NAbs)
cleaned_celltypes <- bub_df2$comp_celltype
for(tu in timepoints_to_use){
  cleaned_celltypes <- str_remove_all(cleaned_celltypes,paste0("__",str_replace_all(tu,"-","vs")))
}

bub_df2$celltype <- cleaned_celltypes




  #### if selected tab combos, keep cell pops on the x-axis

      
         bub_df_bothordered <- bub_df2 %>%
           #### trim to desired assays
           dplyr::filter(assay %in% assay_arrangement) %>%
      ### remove factor levels for assay
      dplyr::mutate(assay = factor(as.character(assay),assay_arrangement),
                    ### also breakout for celltype
                    celltype_only = str_after_first(celltype,"__"),
                    compartment = factor(compartment, c("blood","LN"))) %>%   
      dplyr::arrange(assay) %>%
      #### trim to barbara_order celltypes
      dplyr::filter(celltype %in% barbara_order) %>%
      ## lock in factor levels for assay (rev for y axis since want it to be reversed)
      dplyr::mutate(assay = factor(assay, rev(unique(assay))),
                    ### now get celltype order
                    celltype = factor(celltype, intersect(barbara_order,unique(celltype)))) %>%
           #### now arrange by timepoint to be able to see bulls eyes
           dplyr::arrange(desc(timepoint))
    
 
      ###### reversed axes!!!
    bba <- 
      ggplot(bub_df_bothordered, aes(x = celltype, y = assay))+
      geom_point(aes(size = timepoint, color = timepoint))+
      #scale_alpha_manual(values = alpha_values)+
      scale_size_manual(values = c(3,6,10))+
      scale_color_manual(values = c("#262626","#7F7F7F","#BFBFBF"))+
      theme_bw()+
      ### make legend symbol bigger & fix color 
      # guides(color = guide_legend(override.aes = list(size = 5)),
      #        size = guide_legend(override.aes = list(color = leg_color)))+
      ## remove gridlines
      theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            ## rotate axis labels
            axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
  
  
    plot(bba)

  cat("\n")

    cat(' \n \n') ### this is the key!

      ### last plot in loop doesn't appear unless add this
  
    
 
 
  cat("\n")

    cat(' \n \n') ### this is the key!

    ###### need to fix why this plots in Rstudio but not renedered

Fig 7E

Spearman correlation between MSD D2vsD1 and CD8 Tfh populations

barbara_order <- c("blood__CD8_Tfh_Ki67pos",
                   "blood__CD8_Tfh_GzmBpos",
                   
                   
                   "LN__CD8_Tfh_Ki67pos",
                   "LN__CD8_Tfh_GzmBpos")
 

## only for the Intervals
timepoints_to_use <- paste0(c("D3-D1","D8-D1","D22-D1"),"_",str_after_first(params$log2_or_raw,"value_"),"FC")
TP_name <- "Intervals"

corr_tag_name <- "CorrVsMSD" 
raw_or_adj <- "raw"
cor_method <- "spearman"
trim_pval_level <- 0.05
all_or_selected <- "selected"
assay_arrangement <- c("CCL2.MCP.1","CCL3.MIP.1a","CCL4.MIP.1b","CCL8.MCP.2",
                       "CXCL10.IP.10","CXCL11.ITAC","IL.15","M.CSF",
                       "CCL26.Eotaxin.3","CCL19.MIP.3b","FLT3L","CCL13.MCP.4","CXCL12.SDF.1a",
                       "CX3CL1.Fractalkine") ### set pos/neg/both
directions_to_use <- c("pos")
assays_to_exclude <- c("CCL27.CTACK","IL.5")

raw p < 0.05

#### make df for plotting 
df_long1 <- fullcombo_df_long %>%
  ### generate dir column
  dplyr::mutate(dir = ifelse(p < trim_pval_level,
                             ifelse(corr_coef > 0, "pos","neg"),
                             "notSig"),
                ### break out compartment
                compartment = str_before_first(celltype,"__"),
                celltype = str_after_first(celltype,"__"),
                celltype_assay = str_after_first(celltype_assay,"__")) %>%
  ### remove the corr and p vals for pivoting
  dplyr::select(-c(corr_coef,p)) 



###### set up inital dataframe
## trim to only sig
bub_df2 <- df_long1[which(df_long1$dir %in% directions_to_use ),] %>%
  ### also remove any analytes_to_exclude
  dplyr::filter(!assay %in% assays_to_exclude) %>%
  dplyr::mutate(timepoint = factor(str_before_first(timepoint,"_"),str_replace_all(str_before_first(timepoints_to_use,"_"),"-","vs")),
                dir = factor(dir, directions_to_use),
                ##### add compartment to celltype
                comp_celltype = paste0(compartment,"__",celltype)) 


#### remove timepoint suffix from celltype (only needed for vs Abs/NAbs)
cleaned_celltypes <- bub_df2$comp_celltype
for(tu in timepoints_to_use){
  cleaned_celltypes <- str_remove_all(cleaned_celltypes,paste0("__",str_replace_all(tu,"-","vs")))
}

bub_df2$celltype <- cleaned_celltypes




  #### if selected tab combos, keep cell pops on the x-axis

      
         bub_df_bothordered <- bub_df2 %>%
           #### trim to desired assays
           dplyr::filter(assay %in% assay_arrangement) %>%
      ### remove factor levels for assay
      dplyr::mutate(assay = factor(as.character(assay),assay_arrangement),
                    ### also breakout for celltype
                    celltype_only = str_after_first(celltype,"__"),
                    compartment = factor(compartment, c("blood","LN"))) %>%   
      dplyr::arrange(assay) %>%
      #### trim to barbara_order celltypes
      dplyr::filter(celltype %in% barbara_order) %>%
      ## lock in factor levels for assay (rev for y axis since want it to be reversed)
      dplyr::mutate(assay = factor(assay, rev(unique(assay))),
                    ### now get celltype order
                    celltype = factor(celltype, intersect(barbara_order,unique(celltype)))) %>%
           #### now arrange by timepoint to be able to see bulls eyes
           dplyr::arrange(desc(timepoint))
    
 
      ###### reversed axes!!!
    bba <- 
      ggplot(bub_df_bothordered, aes(x = celltype, y = assay))+
      geom_point(aes(size = timepoint, color = timepoint))+
      #scale_alpha_manual(values = alpha_values)+
      scale_size_manual(values = c(3,6,10))+
      scale_color_manual(values = c("#262626","#7F7F7F","#BFBFBF"))+
      theme_bw()+
      ### make legend symbol bigger & fix color 
      # guides(color = guide_legend(override.aes = list(size = 5)),
      #        size = guide_legend(override.aes = list(color = leg_color)))+
      ## remove gridlines
      theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            ## rotate axis labels
            axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
  
  
    plot(bba)

  cat("\n")

    cat(' \n \n') ### this is the key!

      ### last plot in loop doesn't appear unless add this
  
    
 
 
  cat("\n")

    cat(' \n \n') ### this is the key!

    ###### need to fix why this plots in Rstudio but not renedered

Figure 8

AIM assay

##### set prior for log2 (0 for MSD, 1e-4 for cell pops)
prior_to_use <- 1e-4
curr_figure_name <- "Fig8"

##### select data
data_to_use <- list_of_datasets[["AIM"]] %>%
  ### remove fig name from colnames
  `colnames<-`(c(str_remove_all(colnames(.),paste0(curr_figure_name,"__")))) 

 
### generate dataframe: rows = samples, columns = variables and group for comparison (= TimePoint)

curr_df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  dplyr::mutate(log2_value = log2(raw_value + as.double(prior_to_use)),
                raw_value = NULL,
                ### combine variable + timepoint
                MacaqueID = str_before_first(sampleName,"__"),
                TimePoint = str_after_first(sampleName,"__"), ### includes the vaccine
                sampleName = NULL
                ) 


### select comparisons to make: time1 - time2
TimePoints_to_compare <- data.frame(
  "time1"=c("Spike"),
  "time2"=c("medium")) %>%
  dplyr::mutate(name = paste0(time1,"-",time2))




## set raw or adj p-val
raw_or_adj <- "_pval" #"_adjpval"

### To name output files
contrast_name <- "Fig8ABDE"

### for scatterplots
plot_sig_only <- "no"
num_columns_for_plot <- 3

#### try glm

curr_df <- curr_df_long %>%
  ### make wider into dataset format
  tidyr::pivot_wider(names_from = "variable", values_from = "log2_value") 



  ## prep pvaldf
  p_value_df2 <- data.frame("rowname" = unique(curr_df_long$variable))
  
  for (npt in 1:dim(TimePoints_to_compare)[1]){
    
    ### select animals with both timepoints
    animal_df <- as.data.frame.matrix(table(curr_df$MacaqueID, curr_df$TimePoint)) %>%
      dplyr::select(!!!rlang::syms(c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt]))) %>%
      ## rename to generic
      dplyr::rename(time2=1, time1=2) %>%
      dplyr::mutate(combo = (time2+time1)) %>%
      dplyr::filter(combo ==2) %>%
      rownames_to_column()
 
      
     #trim to only the 2 TimePoints, will remove vars without these TimePoint
      curr_df2 <- curr_df %>% 
        # only animals with both timepoints
        dplyr::filter(MacaqueID %in% animal_df$rowname) %>%

        # only rows with time1 or time2
        dplyr::filter(TimePoint %in% c(TimePoints_to_compare$time1[npt],TimePoints_to_compare$time2[npt])) %>% 
        ## remove variables with all na
        dplyr::select(where(function(x) !all(is.na(x)))) %>%
        #### prep for glm
        dplyr::mutate(TimePoint_f = factor(TimePoint, c(TimePoints_to_compare$time2[npt],TimePoints_to_compare$time1[npt])),
                      ID_f = factor(MacaqueID)) %>%
        dplyr::arrange(ID_f)
      
         ### select variables with both TimePoints
    curr_vars <- colnames(curr_df2)[-which(colnames(curr_df2) %in% c("MacaqueID","TimePoint","TimePoint_f","ID_f"))]

      

      ######## get rid of columns that don't have data for both timepoints
           ######### only for columns that are not all NAs
      NA_columns_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),] %>% keep(~all(is.na(.x))) %>% names #https://www.statology.org/r-find-columns-with-all-na/
      NA_columns_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),] %>% keep(~all(is.na(.x))) %>% names   
      ####### also remove columns with all zeros for both timepoints
      allZero_columns <- which(colSums(curr_df2[,intersect(colnames(curr_df2),curr_vars)]) == 0) %>% names
      
      ####### finally, remove columns with no animals that have both timepoints
      noOverlap_columns <- c()
      for(curr_v in intersect(colnames(curr_df2),curr_vars)){
        animals_tp1 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time1[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        animals_tp2 <- curr_df2[which(curr_df2$TimePoint == TimePoints_to_compare$time2[npt]),c("MacaqueID",curr_v)] %>%
          dplyr::filter(!is.na(!!rlang::sym(curr_v)))
        if(length(intersect(animals_tp1$MacaqueID,animals_tp2$MacaqueID)) == 0){
          noOverlap_columns <- c(noOverlap_columns,curr_v)
        }else{
          noOverlap_columns <- noOverlap_columns
        }
        
      }
      
      nonNA_columns <- setdiff(setdiff(setdiff(setdiff(curr_vars, NA_columns_tp1), NA_columns_tp2),allZero_columns),noOverlap_columns)
      
      ### if there are any columns with data for both TimePoints, compute for those columns
      if(length(nonNA_columns)>0){
        new_variable_mat_prep <- data.frame(as.matrix(data.matrix(curr_df2[,nonNA_columns])))
          
        colnames(new_variable_mat_prep) <- nonNA_columns
        new_variable_mat <- new_variable_mat_prep %>%
          ### remove if all have identical value
          select(where(~n_distinct(.) > 1)) %>%
          as.matrix(.)
        
        ##### glm not giving sensical for BAMA ---> need to log10
        curr_p_raw <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Pr(>|W|)'])
        curr_p_adj <- p.adjust(curr_p_raw, method = 'BH')
        curr_est <- apply(new_variable_mat, 2, function(x) coef(summary(geeglm(scale(x) ~ TimePoint_f, id=ID_f, data = curr_df2, family="gaussian", corstr="exchangeable")))[2,'Estimate'])

        #### no singular errors = good!

            
        
        curr_p_df <- data.frame(curr_est, curr_p_raw, curr_p_adj) %>%
          rownames_to_column()
        
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
        
        
        ###### if there are no columns with nonNA data for this TimePoint, then add nothing
      }else{
        curr_p_df <- data.frame("rowname"=curr_vars,
                                "est"=NA, 
                                "pval"=NA,
                                "adjpval"=NA) 
        colnames(curr_p_df) <- c("rowname",
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_log2FC"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_pval"),
                                 paste0(TimePoints_to_compare$time1[npt],"-",
                                        TimePoints_to_compare$time2[npt],"_adjpval"))
        
      }
      
      
      p_value_df2 <- dplyr::full_join(p_value_df2, curr_p_df, by="rowname")
      
      
  
  }
  


### save stats to list
list_of_stats[[paste0(contrast_name)]] <- p_value_df2

Fig 8A

varcomp_to_plot <- colnames(list_of_datasets[["AIM"]])[grepl("panel8A",colnames(list_of_datasets[["AIM"]]))] %>%
  str_remove_all("Fig8__")

num_columns_for_plot <- 1

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                treat = factor(str_after_first(sampleName,"__"),c("medium","Spike")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = treat, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 6), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 8B

varcomp_to_plot <- colnames(list_of_datasets[["AIM"]])[grepl("panel8B",colnames(list_of_datasets[["AIM"]]))]%>%
  str_remove_all("Fig8__")

num_columns_for_plot <- 1

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                treat = factor(str_after_first(sampleName,"__"),c("medium","Spike")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = treat, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 6), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 8C

varcomp_to_plot <- colnames(list_of_datasets[["AIM"]])[grepl("panel8C",colnames(list_of_datasets[["AIM"]]))]%>%
  str_remove_all("Fig8__")

num_columns_for_plot <- 1

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                treat = factor(str_after_first(sampleName,"__"),c("medium","Spike")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = treat, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 6), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 8D

varcomp_to_plot <- colnames(list_of_datasets[["AIM"]])[grepl("panel8D",colnames(list_of_datasets[["AIM"]]))]%>%
  str_remove_all("Fig8__")

num_columns_for_plot <- 3

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                treat = factor(str_after_first(sampleName,"__"),c("medium","Spike")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = treat, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 6), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Fig 8E

varcomp_to_plot <- colnames(list_of_datasets[["AIM"]])[grepl("panel8E",colnames(list_of_datasets[["AIM"]]))]%>%
  str_remove_all("Fig8__")

num_columns_for_plot <- 3

df_long <- data_to_use %>%
  dplyr::select(-c(MacaqueID,treat)) %>%
  ### make longer to log2 
  tidyr::pivot_longer(cols=!sampleName, names_to = "variable", values_to = "raw_value") %>%
  ### trim to only the vars in this panel
  dplyr::filter(variable %in% varcomp_to_plot) %>%
  ### break out compartment and variable
  dplyr::mutate(comp_var = variable,
                compartment = factor(str_before_first(variable,"__"),c("blood","LN")),
                variable = factor(str_after_first(variable,"__"),varcomp_to_plot),
                treat = factor(str_after_first(sampleName,"__"),c("medium","Spike")),
                MacaqueID = str_before_first(sampleName,"__"),
                sampleName = NULL) %>%
  dplyr::arrange(compartment, variable)

### set up p-values: rowname (comp_var), contrast, log2FC, pval, adjpval
pval_df_trimmed <- list_of_stats[[paste0(contrast_name)]] %>%
  ###make longer
  tidyr::pivot_longer(cols=!rowname, names_to = "contrast_test", values_to = "value") %>%
  dplyr::mutate(test = str_after_first(contrast_test,"_"),
                contrast = str_before_first(contrast_test,"_"),
                contrast_test = NULL) %>%
  tidyr::pivot_wider(names_from = "test",values_from = "value") %>%
  data.frame()

### for each comp_var
list_of_pvals <- list()
for (i in unique(pval_df_trimmed$rowname)){
  pval_df <- tibble::tribble(~group1, ~group2, ~p, ~est, ~dir)
  for (g in 1:nrow(TimePoints_to_compare)){
    
    #### only if have the timepoints
    if(dim(pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),])[1] > 0){
        ### tribble
    curr_df <- tibble::tribble(~group1, ~group2, ~p, ~est,
                               TimePoints_to_compare[g,"time1"],TimePoints_to_compare[g,"time2"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"pval"],
                               pval_df_trimmed[which(pval_df_trimmed$rowname == i & pval_df_trimmed$contrast == TimePoints_to_compare[g,"name"]),"log2FC"]
                               ) %>%
      ### add dir
      dplyr::mutate(dir = ifelse(p < params$pCutoff, 
                                   ifelse(est > 0, "up","down"),
                                   "not_sig"),
                    ### set sig figs
                    p = signif(p,3)) 
      
      
    
    pval_df <- add_row(pval_df, curr_df)

    }else{
      pval_df <- pval_df
    }
     }
  list_of_pvals[[paste0(i)]] <- pval_df
}


list_of_plots <- list()
## for each comp var
for(i in unique(df_long$comp_var)){
 phen_to_plot <- df_long %>%
    dplyr::filter(comp_var == i)
  
  # cat("##### ",paste0(i," "),"\n")
  # cat("\n")
  max_y <- max(phen_to_plot$raw_value, na.rm=TRUE)
  min_y <- min(phen_to_plot$raw_value, na.rm=TRUE)
  range_y <- max_y-min_y
  num_contrasts <- ifelse(plot_sig_only =="no",
                          nrow(TimePoints_to_compare),
                          length(which(list_of_pvals[[paste0(i)]][,"p"] < 0.05))
  )
  maxy_y_to_plot <- ifelse(plot_sig_only =="no",
                                  max_y*(1+0.1*(num_contrasts+1)),
                                  max_y*(1+0.1*(num_contrasts+1))
                                         
                                         )
  
  # color_df <- data.frame("variable"=levels(phen_to_plot$variable),
  #                        
  #                        "fill"=c("#929292","#008F00","#008EF1"))

  #### base plot
  p <- ggplot(phen_to_plot, aes(x = treat, y = raw_value)) + 
#        scale_color_manual(values=c(color_df$fill[which(color_df$variable == str_after_first(i,"__"))])) +  

     geom_point(#aes(color=variable), 
                   size = 2, shape=16, alpha=0.8, cex = 3)+
    geom_line(aes(group = MacaqueID), color = "grey30")+
  scale_y_continuous(limits = c(min_y*(0.8),maxy_y_to_plot)
                     )+
    labs(y = paste0(i))+
  theme_bw()+
        theme(legend.position="none", 
            panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
            panel.border = element_blank(), axis.line = element_line(colour = "black"),
            axis.title.y = element_text(size = 6), ### make smaller to fit
            axis.title.x = element_blank())
  
  
  
  box_pval_cut <- 0.05
  
  ##### add p values
  
  ### if plotting all p
if(plot_sig_only == "no"){

  
  for(l in 1:num_contrasts){
     p <- p +
        stat_pvalue_manual(list_of_pvals[[i]][l,], label = "p",size = 2.5,
                           y.position = max_y+0.1*l*range_y, 
                           bracket.size = 0.2, tip.length = 0)
  }
  
}else{
 
  # if there are any sig contrasts
  if(num_contrasts > 0){
     ## trim to sig contrasts
        sig_pvals <- list_of_pvals[[i]][which(list_of_pvals[[i]][,"p"]  < 0.05),] %>%
          ### add name
        dplyr::left_join(TimePoints_to_compare, by =c("group1","group2"))
        
    for(l in 1:nrow(sig_pvals)){
     # if(pval_combined[which(rownames(pval_combined) == i),paste0(sig_comp_pairs_df$name[l])] < as.double(box_pval_cut)){
        p <- p +
          stat_pvalue_manual(sig_pvals[l,], label = "p",size = 2.5,
                             y.position = max_y+0.1*l*range_y, bracket.size = 0.2, tip.length = 0)
                             
    #  }else{
    #    p <- p
    #  }
        }
      
      
  }else{
    p <- p
  }   

  
}
    
 
  # plot(p)
  list_of_plots[[i]] <- p
  
  
  
  cat("\n")
  cat(' \n \n') ### this is the key!
  
}

do.call("grid.arrange",c(list_of_plots, ncol=num_columns_for_plot))

Data Tables

Compiled raw data

##save
openxlsx::write.xlsx(list_of_datasets,file.path(outputTABLES_dir,paste0("list_of_RAW_datasets.xlsx")))

#### add with button
xfun::embed_file(file.path(outputTABLES_dir,paste0("list_of_RAW_datasets.xlsx")), text = "Excel spreadsheet of RAW datasets")

Excel spreadsheet of RAW datasets

log2 FC

##save
openxlsx::write.xlsx(list_of_diff_counts,file.path(outputTABLES_dir,paste0("list_of_log2FC.xlsx")))

#### add with button
xfun::embed_file(file.path(outputTABLES_dir,paste0("list_of_log2FC.xlsx")), text = "Excel spreadsheet of log2 FC")

Excel spreadsheet of log2 FC

Stats

##save
openxlsx::write.xlsx(list_of_stats,file.path(outputTABLES_dir,paste0("list_of_stats.xlsx")))

#### add with button
xfun::embed_file(file.path(outputTABLES_dir,paste0("list_of_stats.xlsx")), text = "Excel spreadsheet of stats")

Excel spreadsheet of stats

Session Info

sessionInfo()

## R version 4.4.1 (2024-06-14)
## Platform: aarch64-apple-darwin20
## Running under: macOS Sonoma 14.6.1
## 
## Matrix products: default
## BLAS:   /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/lib/libRblas.0.dylib 
## LAPACK: /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.12.0
## 
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
## 
## time zone: America/New_York
## tzcode source: internal
## 
## attached base packages:
## [1] stats4    stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
##  [1] lmerTest_3.1-3         lme4_1.1-35.3          Matrix_1.7-0          
##  [4] geepack_1.3.10         ggalluvial_0.12.5      dendextend_1.17.1     
##  [7] flextable_0.9.6        corrplot_0.92          edgeR_4.2.0           
## [10] limma_3.60.0           EnhancedVolcano_1.22.0 colorspace_2.1-0      
## [13] RColorBrewer_1.1-3     dendsort_0.3.4         strex_2.0.0           
## [16] ggplotify_0.1.2        plotly_4.10.4          ggrepel_0.9.5         
## [19] ggbeeswarm_0.7.2       downloadthis_0.3.3     DT_0.33               
## [22] gridExtra_2.3          org.Hs.eg.db_3.19.1    AnnotationDbi_1.66.0  
## [25] IRanges_2.38.0         S4Vectors_0.42.0       Biobase_2.64.0        
## [28] BiocGenerics_0.50.0    gtsummary_1.7.2        kableExtra_1.4.0      
## [31] broom_1.0.5            knitr_1.46             enrichR_3.2           
## [34] eulerr_7.0.2           ggpubr_0.6.0           pheatmap_1.0.12       
## [37] cowplot_1.1.3          lubridate_1.9.3        forcats_1.0.0         
## [40] stringr_1.5.1          dplyr_1.1.4            purrr_1.0.2           
## [43] readr_2.1.5            tidyr_1.3.1            tibble_3.2.1          
## [46] ggplot2_3.5.1          tidyverse_2.0.0        readxl_1.4.3          
## [49] openxlsx_4.2.5.2      
## 
## loaded via a namespace (and not attached):
##   [1] splines_4.4.1           later_1.3.2             cellranger_1.1.0       
##   [4] polyclip_1.10-6         janitor_2.2.0           lifecycle_1.0.4        
##   [7] rstatix_0.7.2           lattice_0.22-6          MASS_7.3-60.2          
##  [10] crosstalk_1.2.1         backports_1.4.1         magrittr_2.0.3         
##  [13] sass_0.4.9              rmarkdown_2.26          jquerylib_0.1.4        
##  [16] yaml_2.3.8              httpuv_1.6.15           zip_2.3.1              
##  [19] askpass_1.2.0           DBI_1.2.2               minqa_1.2.6            
##  [22] abind_1.4-5             zlibbioc_1.50.0         yulab.utils_0.1.4      
##  [25] WriteXLS_6.5.0          tweenr_2.0.3            gdtools_0.3.7          
##  [28] GenomeInfoDbData_1.2.12 crul_1.4.2              svglite_2.1.3          
##  [31] ggforce_0.4.2           xml2_1.3.6              tidyselect_1.2.1       
##  [34] farver_2.1.2            httpcode_0.3.0          UCSC.utils_1.0.0       
##  [37] viridis_0.6.5           broom.helpers_1.15.0    jsonlite_1.8.8         
##  [40] systemfonts_1.0.6       tools_4.4.1             ragg_1.3.1             
##  [43] Rcpp_1.0.12             glue_1.7.0              xfun_0.43              
##  [46] GenomeInfoDb_1.40.0     withr_3.0.0             numDeriv_2016.8-1.1    
##  [49] fastmap_1.1.1           boot_1.3-30             fansi_1.0.6            
##  [52] openssl_2.1.2           digest_0.6.35           timechange_0.3.0       
##  [55] R6_2.5.1                mime_0.12               gridGraphics_0.5-1     
##  [58] textshaping_0.3.7       RSQLite_2.3.6           utf8_1.2.4             
##  [61] generics_0.1.3          fontLiberation_0.1.0    data.table_1.15.4      
##  [64] httr_1.4.7              htmlwidgets_1.6.4       pkgconfig_2.0.3        
##  [67] gtable_0.3.5            blob_1.2.4              XVector_0.44.0         
##  [70] htmltools_0.5.8.1       fontBitstreamVera_0.1.1 carData_3.0-5          
##  [73] scales_1.3.0            png_0.1-8               snakecase_0.11.1       
##  [76] rstudioapi_0.16.0       tzdb_0.4.0              rjson_0.2.21           
##  [79] uuid_1.2-0              checkmate_2.3.1         nlme_3.1-164           
##  [82] curl_5.2.1              nloptr_2.0.3            cachem_1.0.8           
##  [85] vipor_0.4.7             pillar_1.9.0            grid_4.4.1             
##  [88] vctrs_0.6.5             promises_1.3.0          car_3.1-2              
##  [91] xtable_1.8-4            beeswarm_0.4.0          evaluate_0.23          
##  [94] cli_3.6.2               locfit_1.5-9.9          compiler_4.4.1         
##  [97] rlang_1.1.4             crayon_1.5.2            ggsignif_0.6.4         
## [100] labeling_0.4.3          fs_1.6.4                stringi_1.8.4          
## [103] viridisLite_0.4.2       munsell_0.5.1           Biostrings_2.72.0      
## [106] lazyeval_0.2.2          fontquiver_0.2.1        hms_1.1.3              
## [109] bit64_4.0.5             gfonts_0.2.0            KEGGREST_1.44.0        
## [112] statmod_1.5.0           shiny_1.8.1.1           highr_0.10             
## [115] gt_0.10.1               memoise_2.0.1           bslib_0.7.0            
## [118] bit_4.0.5               officer_0.6.6